Frank da Cruz
The Kermit Project, Columbia University
As of:
C-Kermit 8.0.206 24 Oct 2002
This page last updated:
Sat Dec 7 11:58:57 2002
(New York USA Time)
IF YOU ARE READING A PLAIN-TEXT version of this document, note it is a plain-text dump of a Web page. You can visit the original (and possibly more up-to-date) Web page here:
http://www.columbia.edu/kermit/ckubwr.html
Since the material in this file has been accumulating since 1985, some (much) of it might be dated. Feedback from experts on particular OS's and platforms is always welcome.
[ C-Kermit ] [ Installation Instructions ] [ TUTORIAL ]
Quick Links: [ Linux ] [ *BSD ] [ AIX ] [ HP-UX ] [ Solaris ] [ SCO ] [ DEC/Compaq ]
SECTION CONTENTS
1.1. Documentation 1.2. Technical Support 1.3. The Year 2000 1.4. The Euro
THIS IS WHAT USED TO BE CALLED the "beware file" for the Unix version of C-Kermit, previously distributed as ckubwr.txt and, before that, as ckuker.bwr, after the fashion of old Digital Equipment Corporation (DEC) software releases that came with release notes (describing what had changed) and a "beware file" listing known bugs, limitations, "non-goals", and things to watch out for. The C-Kermit beware file has been accumulating since 1985, and it applies to many different hardware platforms and operating systems, and many versions of them, so it is quite large. Prior to C-Kermit 8.0, it was distributed only in plain-text format. Now it is available as a Web document with links, internal cross references, and so on, to make it easier to use.
This document applies to Unix C-Kermit in general, as well as to specific Unix variations like Linux, AIX, HP-UX, Solaris, and so on, and should be read in conjunction with the platform-independent C-Kermit beware file, which contains similar information, but applying to all versions of C-Kermit (VMS, Windows, OS/2, AOS/VS, VOS, etc, as well as to Unix).
There is much in this document that is (only) of historical interest. The navigation links should help you skip directly to the sections that are relevant to you. Numerous offsite Web links are supposed to lead to further information but, as you know, Web links go stale frequently and without warning. If you can supply additional, corrected, updated, or better Web links, please feel free to let us know.
C-Kermit 6.0 is documented in the book Using C-Kermit, Second Edition, by Frank da Cruz and Christine M. Gianone, Digital Press, Burlington, MA, USA, ISBN 1-55558-164-1 (1997), 622 pages. This remains the definitive C-Kermit documentation. Until the third edition is published (sorry, there is no firm timeframe for this), please also refer to:
For information on how to get technical support, please visit:
http://www.columbia.edu/kermit/support.html
The Unix version of C-Kermit, release 6.0 and later, is "Year 2000 compliant", but only if the underlying operating system is too. Contact your Unix operating system vendor to find out which operating system versions, patches, hardware, and/or updates are required. (Quite a few old Unixes are still in operation in the new millenium, but with their date set 28 years in the past so at least the non-year parts of the calendar are correct.)
As of C-Kermit 6.0 (6 September 1996), post-millenium file dates are recognized, transmitted, received, and reproduced correctly during the file transfer process in C-Kermit's File Attribute packets. If post-millenium dates are not processed correctly on the other end, file transfer still takes place, but the modification or creation date of the received file might be incorrect. The only exception would be if the "file collision update" feature is being used to prevent unnecessary transfer of files that have not changed since the last time a transfer took place; in this case, a file might be transferred unnecessarily, or it might not be transferred when it should have been. Correct operation of the update feature depends on both Kermit programs having the correct date and time.
Of secondary importance are the time stamps in the transaction and/or debug logs, and the date-related script programming constructs, such as \v(date), \v(ndate), \v(day), \v(nday), and perhaps also the time-related ones, \v(time) and \v(ntime), insofar as they might be affected by the date. The \v(ndate) is a numeric-format date of the form yyyymmdd, suitable for both lexical and numeric comparison and sorting: e.g. 19970208 or 20011231. If the underlying operating system returns the correct date information, these variables will have the proper values. If not, then scripts that make decisions based on these variables might not operate correctly.
Most date-related code is based upon the C Library asctime() string, which always has a four-digit year. In Unix, the one bit of code in C-Kermit that is an exception to this rule is several calls to localtime(), which returns a pointer to a tm struct, in which the year is presumed to be expressed as "years since 1900". The code depends on this assumption. Any platforms that violate it will need special coding. As of this writing, no such platforms are known.
Command and script programming functions that deal with dates use C-Kermit specific code that always uses full years.
C-Kermit 7.0 and later support Unicode (ISO 10646), ISO 8859-15 Latin Alphabet 9, PC Code Page 858, Windows Code Pages 1250 and 1251, and perhaps other character sets, that encode the Euro symbol, and can translate among them as long as no intermediate character-set is involved that does not include the Euro.
It is often dangerous to run a binary C-Kermit (or any other) program built on a different computer. Particularly if that computer had a different C compiler, libraries, operating system version, processor features, etc, and especially if the program was built with shared libraries, because as soon as you update the libraries on your system, they no longer match the ones referenced in the binary, and the binary might refuse to load when you run it, in which case you'll see error messages similar to:
Could not load program kermit Member shr4.o not found or file not an archive Could not load library libcurses.a[shr4.o] Error was: No such file or directory
(These samples are from AIX.) To avoid this problem, we try to build C-Kermit with statically linked libraries whenever we can, but this is increasingly impossible as shared libraries become the norm.
It is often OK to run a binary built on an earlier OS version, but it is rarely possible (or safe) to run a binary built on a later one, for example to run a binary built under Solaris 8 on Solaris 2.6. Sometimes even the OS-or-library patch/ECO level makes a difference.
A particularly insidious problem occurs when a binary was built on a version of the OS that has patches from the vendor (e.g. to libraries); in many cases you won't be able to run such a binary on an unpatched version of the same platform.
When in doubt, build C-Kermit from the source code on the computer where it is to be run (if possible!). If not, ask us for a binary specific to your configuration. We might have one, and if we don't, we might be able to find somebody who will build one for you.
SECTION CONTENTS
3.0. C-KERMIT ON PC-BASED UNIXES
3.1. C-KERMIT AND AIX
3.2. C-KERMIT AND HP-UX
3.3. C-KERMIT AND LINUX
3.4. C-KERMIT AND NEXTSTEP
3.5. C-KERMIT AND QNX
3.6. C-KERMIT AND SCO
3.7. C-KERMIT AND SOLARIS
3.8. C-KERMIT AND SUNOS
3.9. C-KERMIT AND ULTRIX
3.10. C-KERMIT AND UNIXWARE
3.11. C-KERMIT AND APOLLO SR10
3.12. C-KERMIT AND TANDY XENIX 3.0
3.13. C-KERMIT AND OSF/1 (DIGITAL UNIX) (TRU64 UNIX)
3.14. C-KERMIT AND SGI IRIX
3.15. C-KERMIT AND THE BEBOX
3.16. C-KERMIT AND DG/UX
3.17. C-KERMIT AND SEQUENT DYNIX
3.18. C-KERMIT AND {FREE,OPEN,NET}BSD
3.19. C-KERMIT AND MAC OS X (Rhapsody, Darwin)
3.20. C-KERMIT AND COHERENT
The following sections apply to specific Unix versions. Most of them contain references to FAQs (Frequently Asked Questions), but these tend to be ephemeral. For possibly more current information see:
http://www.faqs.org
One thread that runs through many of them, and implicitly perhaps through all, concerns the problems that occur when trying to dial out on a serial device that is (also) enabled for dialing in. The "solutions" to this problem are many, varied, diverse, and usually gross, involving configuring the device for bidirectional use. This is done in a highly OS-dependent and often obscure manner, and the effects (good or evil) are also highly dependent on the particular OS (and getty variety, etc). Many examples are given in the OS-specific sections below.
An important point to keep in mind is that C-Kermit is a cross-platform, portable software program. It was not designed specifically and only for your particular Unix version, or for that matter, for Unix in particular at all. It also runs on VMS, AOS/VS, VOS, and other non-Unix platforms. All the Unix versions of C-Kermit share common i/o modules, with compile-time #ifdef constructions used to account for the differences among the many Unix products and releases. If you think that C-Kermit is behaving badly or missing something on your particular Unix version, you might be right -- we can't claim to be expert in hundreds of different OS / version / hardware / library combinations. If you're a programmer, take a look at the source code and send us your suggested fixes or changes. Or else just send us a report about what seems to be wrong and we'll see what we can do.
SECTION CONTENTS
3.0.1. Interrupt Conflicts 3.0.2. Windows-Specific Hardware 3.0.3. Modems 3.0.4. Character Sets 3.0.5. Keyboard, Screen, and Mouse Access 3.0.6. Laptops
PCs are not the best platform for real operating systems like Unix. The architecture suffers from numerous deficiencies, not the least of which is the stiflingly small number of hardware interrupts (either 7 or 15, many of which are preallocated). Thus adding devices, using multiple serial ports, etc, is always a challenge and often a nightmare. The free-for-all nature of the PC market and the lack of standards combined with the diversity of Unix OS versions make it difficult to find drivers for any particular device on any particular version of Unix.
Of special interest to Kermit users is the fact that there is no standard provision in the PC architecture for more than 2 communication (serial) ports. COM3 and COM4 (or higher) will not work unless you (a) find out the hardware address and interrupt for each, (b) find out how to provide your Unix version with this information, and (c) actually set up the configuration in the Unix startup files (or whatever other method is used). Watch out for interrupt conflicts, especially when using a serial mouse, and don't expect to be able to use more than two serial ports.
The techniques for resolving interrupt conflicts are different for each operating system (Linux, NetBSD, etc). In general, there is a configuration file somewhere that lists COM ports, something like this:
com0 at isa? port 0x3f8 irq 4 # DOS COM1 com1 at isa? port 0x2f8 irq 3 # DOS COM2
The address and IRQ values in this file must agree with the values in the PC BIOS and with the ports themselves, and there must not be more than one device with the same interrupt. Unfortunately, due to the small number of available interrupts, installing new devices on a PC almost always creates a conflict. Here is a typical tale from a Linux user (Fred Smith) about installing a third serial port:
...problems can come from a number of causes. The one I fought with for some time, and finally conquered, was that my modem is on an add-in serial port, cua3/IRQ5. By default IRQ5 has a very low priority, and does not get enough service in times when the system is busy to prevent losing data. This in turn causes many resends. There are two 'fixes' that I know of, one is to relax hard disk interrupt hogging by using the correct parameter to hdparm, but I don't like that one because the hdparm man page indicates it is risky to use. The other one, the one I used, was to get 'irqtune' and use it to give IRQ5 the highest priority instead of nearly the lowest. Completely cured the problem.
Here's another one from a newsgroup posting:
After much hair pulling, I've discovered why my serial port won't work. Apparently my [PC] has three serial devices (two comm ports and an IR port), of which only two at a time can be active. I looked in the BIOS setup and noticed that the IR port was activated, but didn't realize at the time that this meant that COM2 was thereby de-activated. I turned off the IR port and now the serial port works as advertised.
To complicate matters, the PC platform is becoming increasingly and inexorably Windows-oriented. More and more add-on devices are "Windows only" -- meaning they are incomplete and rely on proprietary Windows-based software drivers to do the jobs that you would expect the device itself to do. PCMCIA, PCI, or "Plug-n-Play" devices are rarely supported on PC-based Unix versions such as SCO; Winmodems, Winprinters, and the like are not supported on any Unix variety (with a few exceptions). The self-proclaimed Microsoft PC 97 (or later) standard only makes matters worse since its only purpose to ensure that PCs are "optimized to run Windows 95 and Windows NT 4.0 and future versions of these operating systems".
With the exception noted (the Lucent modem, perhaps a handful of others by the time you read this), drivers for "Win" devices are available only for Windows, since the Windows market dwarfs that of any particular Unix brand, and for that matter all Unixes (or for that matter, all non-Windows operating systems) combined. If your version of Unix (SCO, Linux, BSDI, FreeBSD, etc) does not support a particular device, then C-Kermit can't use it either. C-Kermit, like any Unix application, must access all devices through drivers and not directly because Unix is a real operating system.
Don't waste time thinking that you, or anybody else, could write a Linux (or other Unix) driver for a Winmodem or other "Win" device. First of all, these devices generally require realtime control, but since Unix is a true multitasking operating system, realtime device control is not possible outside the kernel. Second, the specifications for these devices are secret and proprietary, and each one (and each version of each one) is potentially different. Third, a Winmodem driver would be enormously complex; it would take years to write and debug, by which time it would be obsolete.
A more recent generation of PCs (circa 1999-2000) is marketed as "Legacy Free". One can only speculate what that could mean. Most likely it means it will ONLY run the very latest versions of Windows, and is made exclusively of Winmodems, Winprinters, Winmemory, and Win-CPU-fans (Legacy Free is a concept pioneered by Microsoft).
Before you buy a new PC or add-on equipment, especially serial ports, internal modems, or printers, make sure they are compatible with your version of Unix. This is becoming an ever-greater challenge; only a huge company like Microsoft can afford to be constantly cranking out and/or verifying drivers for the thousands of video boards, sound cards, network adapters, SCSI adapters, buses, etc, that spew forth in an uncontrolled manner from all corners of the world on a daily basis. With very few exceptions, makers of PCs assemble the various components and then verify them only with Windows, which they must do since they are, no doubt, preloading the PC with Windows. To find a modern PC that is capable of running a variety of non-Windows operating systems (e.g. Linux, SCO OpenServer, Unixware, and Solaris) is a formidable challenge requiring careful study of each vendor's "compatibility lists" and precise attention to exact component model numbers and revision levels.
External modems are recommended:
Internal PC modems (even when they are not Winmodems, which is increasingly unlikely in new PCs) are always trouble, especially in Unix. Even when they work for dialing out, they might not work for dialing in, etc. Problems that occur when using an internal modem can almost always be eliminated by switching to an external one. Even when an internal modem is not a Winmodem or Plug-n-Play, it is often a no-name model of unknown quality -- not the sort of thing you want sitting directly on your computer's bus. (Even if it does not cause hardware problems, it probably came without a command list, so no Unix software will know how to control it.) For more about Unix compatible modems, see:
http://www.o2.net/~gromitkc/winmodem.html
Remember that PCs, even now -- more than two decades after they were first introduced -- are not (in general) capable of supporting more than 2 serial devices. Here's a short success story from a recent newsgroup posting: "I have a Diamond SupraSonic II dual modem in my machine. What I had to end up doing is buying a PS/2 mouse and port and install it. Had to get rid of my serial mouse. I also had to disable PnP in my computer bios. I was having IRQ conflicts between my serial mouse and 'com 3'. Both modems work fine for me. My first modem is ttyS0 and my second is ttyS1." Special third-party multiport boards such as DigiBoard are available for certain Unix platforms (typically SCO, maybe Linux) that come with special platform-specific drivers.
PCs generally have PC code pages such as CP437 or CP850, and these are often used by PC-based Unix operating systems, particularly on the console. These are supported directly by C-Kermit's SET FILE CHARACTER-SET and SET TERMINAL CHARACTER-SET commands. Some PC-based Unix versions, such as recent Red Hat Linux releases, might also support Microsoft Windows code pages such as CP1252, or even Latin Alphabet 1 itself (perhaps displayed with CP437 glyphs). (And work is in progress to support Unicode UTF8 in Linux.)
Certain Windows code pages are not supported directly by C-Kermit, but since they are ISO Latin Alphabets with nonstandard "extensions" in the C1 control range, you can substitute the corresponding Latin alphabet (or other character set) in any C-Kermit character-set related commands:
Windows Code Page Substitution CP 1004 Latin-1 CP 1051 HP Roman-8
Other Windows code pages are mostly (or totally) incompatible with their Latin Alphabet counterparts (e.g. CP1250 and Latin-2), and several of these are already supported by C-Kermit 7.0 and later (1250, 1251, and 1252).
Finally, note that as a real operating system, Unix (unlike Windows) does not provide the intimate connection to the PC keyboard, screen, and mouse that you might expect. Unix applications can not "see" the keyboard, and therefore can not be programmed to understand F-keys, Editing keys, Arrow keys, Alt-key combinations, and the like. This is because:
(To be filled in . . .)
SECTION CONTENTS
3.1.1. AIX: General 3.1.2. AIX: Network Connections 3.1.3. AIX: Serial Connections 3.1.4. AIX: File Transfer 3.1.5. AIX: Xterm Key Map
For additional information see:
and/or read the comp.unix.aix newsgroup.
About AIX version numbers: "uname -a" tells the two-digit version number, such as 3.2 or 4.1. The three-digit form can be seen with the "oslevel" command (this information is unavailable at the API level and is reportedly obtained by scanning the installed patch list). Supposedly all three-digit versions within the same two-digit version (e.g. 4.3.1, 4.3.2) are binary compatible; i.e. a binary built on any one of them should run on all others, but who knows. Most AIX advocates tell you that any AIX binary will run on any AIX version greater than or equal to the one under which it was built, but experience with C-Kermit suggests otherwise. It is always best to run a binary built under your exact same AIX version, down to the third decimal place, if possible. Ideally, build it from source code yourself.
File transfers into AIX 4.2 or 4.3 through the AIX Telnet or Rlogin server have been observed to fail (or accumulate huge numbers of correctable errors, or even disconnect the session), when exactly the same kind of transfers into AIX 4.1 work without incident, as do such transfers into all non-AIX platforms on the same kind of connections (with a few exceptions noted elsewhere in this document). AIX 4.3.3 seems to be particularly fragile in this regard; the weakness seems to be in its pseudoterminal (pty) driver. High-speed streaming transfers work perfectly, however, if the AIX Telnet server and pty driver are removed from the picture; e.g, by using "set host * 3000" on AIX.
The problem can be completely cured by replacing the IBM Telnet server with MIT's Kerberos Telnet server -- even if you don't actually the Kerberos part. Diagnosis: AIX pseudoterminals (which are controlled by the Telnet server to give you a login terminal for your session) have quirks that not even IBM knows about. The situation with AIX 5.x is not known, but if it has the same problem, the same cure is available.
Meanwhile, the only remedy when going through the IBM Telnet server is to cut back on Kermit's performance settings until you find a combination that works:
In some cases, severe cutbacks are required, e.g. those implied by the ROBUST command. Also be sure that the AIX C-Kermit on the remote end has "set flow none" (which is the default). NOTE: Maybe this one can also be addressed by starting AIX telnetd with the "-a" option. The situation with SSH connections is not known, but almost certainly the same.
When these problems occur, the system error log contains:
LABEL: TTY_TTYHOG IDENTIFIER: 0873CF9F Type: TEMP Resource Name: pts/1 Description TTYHOG OVER-RUN Failure Causes EXCESSIVE LOAD ON PROCESSOR Recommended Actions REDUCE SYSTEM LOAD. REDUCE SERIAL PORT BAUD RATE
Before leaving the topic of AIX pseudoterminals, it is very likely that Kermit's PTY and SSH commands do not work well either, for the same reason that Telnet connections into AIX don't work well. A brief test with "pty rlogin somehost" got a perfectly usable terminal (CONNECT) session, but file-transfer problems like those just described.
Reportedly, telnet from AIX 4.1-point-something to non-Telnet ports does not work unless the port number is in the /etc/services file; it's not clear from the report whether this is a problem with AIX Telnet (in which case it would not affect Kermit), or with the sockets library (in which case it would). The purported fix is IBM APAR IX61523.
C-Kermit SET HOST or TELNET from one AIX 3.1 (or earlier) system to another won't work right unless you set your local terminal type to something other than AIXTERM. When your terminal type is AIXTERM, AIX TELNET sends two escapes whenever you type one, and the AIX telnet server swallows one of them. This has something to do with the "hft" device. This behavior seems to be removed in AIX 3.2 and later.
In AIX 3, 4, or 5, C-Kermit won't be able to "set line /dev/tty0" (or any other dialout device) if you haven't installed "cu" or "uucp" on your system, because installing these is what creates the UUCP lockfile directory. If SET LINE commands always result in "Sorry, access to lock denied", even when C-Kermit has been given the same owner, group, and permissions as cu:
-r-sr-xr-x 1 uucp uucp 67216 Jul 27 1999 cu
and even when you run it as root, then you must go back and install "cu" from your AIX installation media.
According to IBM's "From Strength to Strength" document (21 April 1998), in AIX 4.2 and later "Async supports speeds on native serial ports up to 115.2kbps". However, no API is documented to achieve serial speeds higher than 38400 bps. Apparently the way to do this -- which might or might not work only on the IBM 128-port multiplexer -- is:
cxma-stty fastbaud /dev/tty0
which, according to "man cxma-stty":
fastbaud Alters the baud rate table, so 50 baud becomes 57600 baud.
-fastbaud Restores the baud rate table, so 57600 baud becomes 50 baud.
Presumably (but not certainly) this extrapolates to 110 "baud" becomes 76800 bps, and 150 becomes 115200 bps. So to use high serial speeds in AIX 4.2 or 4.3, the trick would be to give the "cxma-stty fastbaud" command for the desired tty device before starting Kermit, and then use "set speed 50", "set speed 110", or "set speed 150" to select 56700, 76800, or 115200 bps. It is not known whether cxma-stty requires privilege.
According to one report, "Further investigation with IBM seems to indicate that the only hardware capable of doing this is the 128-port multiplexor with one (or more) of the 16 port breakout cables (Enhanced Remote Async Node 16-Port EIA-232). We are looking at about CDN$4,000 in hardware just to hang a 56kb modem on there. Of course, we can then hang 15 more, if we want. This hardware combo is described to be good to 230.4kbps."
Another report says (quote from AIX newsgroup, March 1999):
The machine type and the adapter determine the speed that one can actually run at. The older microchannel machines have much slower crystal frequencies and may not go beyond 76,800. A feature put into AIX 421 allows one to key in non-POSIX baud rates and if the uart can support that speed, it will get set. this applies also to 43p's and beyond. 115200 is the max for the 43P's native serial port. As crytal frequencies continue to increase, the built-in serial ports speeds will improve. To use 'uucp' or 'ate' at the higher baud rates, configure the port for the desired speed, but set the speed of uucp or ate to 50. Any non-POSIX speeds set in the ttys configuration will the be used. In the case of the 128-port adapters or the ISA 8-port or PCI 8-port adapter, there are only a few higher baud rates.
- Change the port to enable high baud rates:
- B50 for 57600
- B75 for 76800
- B110 for 115200
- B200 for 230000
- chdev -l ttyX -a fastbaud=enable
- For the 128 ports original style rans, only 57600 bps is supported.
- For the new enhanced RANs, up to 230Kbps is supported.
In AIX 2.2.1 on the RT PC with the 8-port multiplexer, SET SPEED 38400 gives 9600 bps, but SET SPEED 19200 gives 19200 (on the built-in S1 port).
Note that some RS/6000s (e.g. the IBM PowerServer 320) have nonstandard rectangular 10-pin serial ports; the DB-25 connector is NOT a serial port; it is a parallel printer port. IBM cables are required for the serial ports, (The IBM RT PC also had rectangular serial ports -- perhaps the same as these, perhaps different.)
If you dial in to AIX through a modem that is connected directly to an AIX port (e.g. on the 128-port multiplexer) and find that data is lost, especially when uploading files to the AIX system (and system error logs report buffer overruns on the port):
http://service.software.ibm.com/rs6000/
Downloads -> Software Fixes -> Download FixDist gets an application for looking up known problems.
Many problems reported with bidirectional terminal lines on AIX 3.2.x on the RS/6000. Workaround: don't use bidirectional terminal lines, or write a shell-script wrapper for Kermit that turns getty off on the line before starting Kermit, or before Kermit attempts to do the SET LINE. (But note: These problems MIGHT be fixed in C-Kermit 6.0 and later.) The commands for turning getty off and on (respectively) are /usr/sbin/pdisable and /usr/sbin/penable.
Evidently AIX 4.3 (I don't know about earlier versions) does not allow open files to be overwritten. This can cause Kermit transfers to fail when FILE COLLISION is OVERWRITE, where they might work on other Unix varieties or earlier AIX versions.
Transfer of binary -- and maybe even text -- files can fail in AIX if the AIX terminal has particular port can have character-set translation done for it by the tty driver. The following advice from a knowledgeable AIX user:
[This feature] has to be checked (and set/cleared) with a separate command, unfortunately stty doesn't handle this. To check:
$ setmaps input map: none installed output map: none installedIf it says anything other than "none installed" for either one, it is likely to cause a problem with kermit. To get rid of installed maps:
$ setmaps -t NOMAPHowever, I seem to recall that with some versions of AIX before 3.2.5, only root could change the setting. I'm not sure what versions - it might have only been under AIX 3.1 that this was true. At least with AIX 3.2.5 an ordinary user can set or clear the maps.
On the same problem, another knowledgeable AIX user says:
The way to get information on the NLS mapping under AIX (3.2.5 anyway) is as follows. From the command line type:
lsattr -l tty# -a imap -a omap -E -HReplace the tty number for the number sign above. This will give a human readable output of the settings that looks like this;
# lsattr -l tty2 -a imap -a omap -E -H attribute value description user_settable
imap none INPUT map file True omap none OUTPUT map file TrueIf you change the -H to a -O, you get output that can easily be processed by another program or a shell script, for example:
# lsattr -l tty2 -a imap -a omap -E -O #imap:omap none:noneTo change the settings from the command line, the chdev command is used with the following syntax.
chdev -l tty# -a imap='none' -a omap='none'Again substituting the appropriate tty port number for the number sign, "none" being the value we want for C-Kermit. Of course, the above can also be changed by using the SMIT utility and selecting devices - tty. (...end quote)
Here is a sample configuration for setting up an xterm keyboard for VT220 or higher terminal emulation on AIX, courtesy of Bruce Momjian, Drexel Hill, PA. Xterm can be started like this:
xterm $XTERMFLAGS +rw +sb +ls $@ -tm 'erase ^? intr ^c' -name vt220 \
-title vt220 -tn xterm-220 "$@" &
---------------------------------------------------------------------------
XTerm*VT100.Translations: #override \n\
<Key>Home: string(0x1b) string("[3~") \n \
<Key>End: string(0x1b) string("[4~") \n
vt220*VT100.Translations: #override \n\
Shift <Key>F1: string("[23~") \n \
Shift <Key>F2: string("[24~") \n \
Shift <Key>F3: string("[25~") \n \
Shift <Key>F4: string("[26~") \n \
Shift <Key>F5: string("[K~") \n \
Shift <Key>F6: string("[31~") \n \
Shift <Key>F7: string("[31~") \n \
Shift <Key>F8: string("[32~") \n \
Shift <Key>F9: string("[33~") \n \
Shift <Key>F10: string("[34~") \n \
Shift <Key>F11: string("[28~") \n \
Shift <Key>F12: string("[29~") \n \
<Key>Print: string(0x1b) string("[32~") \n\
<Key>Cancel: string(0x1b) string("[33~") \n\
<Key>Pause: string(0x1b) string("[34~") \n\
<Key>Insert: string(0x1b) string("[2~") \n\
<Key>Delete: string(0x1b) string("[3~") \n\
<Key>Home: string(0x1b) string("[1~") \n\
<Key>End: string(0x1b) string("[4~") \n\
<Key>Prior: string(0x1b) string("[5~") \n\
<Key>Next: string(0x1b) string("[6~") \n\
<Key>BackSpace: string(0x7f) \n\
<Key>Num_Lock: string(0x1b) string("OP") \n\
<Key>KP_Divide: string(0x1b) string("Ol") \n\
<Key>KP_Multiply: string(0x1b) string("Om") \n\
<Key>KP_Subtract: string(0x1b) string("OS") \n\
<Key>KP_Add: string(0x1b) string("OM") \n\
<Key>KP_Enter: string(0x1b) string("OM") \n\
<Key>KP_Decimal: string(0x1b) string("On") \n\
<Key>KP_0: string(0x1b) string("Op") \n\
<Key>KP_1: string(0x1b) string("Oq") \n\
<Key>KP_2: string(0x1b) string("Or") \n\
<Key>KP_3: string(0x1b) string("Os") \n\
<Key>KP_4: string(0x1b) string("Ot") \n\
<Key>KP_5: string(0x1b) string("Ou") \n\
<Key>KP_6: string(0x1b) string("Ov") \n\
<Key>KP_7: string(0x1b) string("Ow") \n\
<Key>KP_8: string(0x1b) string("Ox") \n\
<Key>KP_9: string(0x1b) string("Oy") \n
! <Key>Up: string(0x1b) string("[A") \n\
! <Key>Down: string(0x1b) string("[B") \n\
! <Key>Right: string(0x1b) string("[C") \n\
! <Key>Left: string(0x1b) string("[D") \n\
*visualBell: true
*saveLines: 1000
*cursesemul: true
*scrollKey: true
*scrollBar: true
SECTION CONTENTS
3.2.0. Common Problems 3.2.1. Building C-Kermit on HP-UX 3.2.2. File Transfer 3.2.3. Dialing Out and UUCP Lockfiles in HP-UX 3.2.4. Notes on Specific HP-UX Releases 3.2.5. HP-UX and X.25
REFERENCES
For further information, read the comp.sys.hp.hpux newsgroup.
C-Kermit is included as part of the HP-UX operating system by contract between Hewlett Packard and Columbia University for HP-UX 10.00 and later. Each level of HP-UX includes a freshly built C-Kermit binary in /bin/kermit, which should work correctly. Binaries built for regular HP-UX may be used on Trusted HP-UX and vice-versa, except for use as IKSD because of the different authentication methods.
Some HP workstations have a BREAK/RESET key. If you hit this key while C-Kermit is running, it might kill or suspend the C-Kermit process. C-Kermit arms itself against these signals, but evidently the BREAK/RESET key is -- at least in some circumstances, on certain HP-UX versions -- too powerful to be caught. (Some report that the first BREAK/RESET shows up as SIGINT and is caught by C-Kermit's former SIGINT handler even when SIGINT is currently set to SIG_IGN; the second kills Kermit; other reports suggest the first BREAK/RESET sends a SIGTSTP (suspend signal) to Kermit, which it catches and suspends itself. You can tell C-Kermit to ignore suspend signals with SET SUSPEND OFF. You can tell C-Kermit to ignore SIGINT with SET COMMAND INTERRUPTION OFF. It is not known whether these commands also grant immunity to the BREAK/RESET key (one report states that with SET SUSPEND OFF, the BREAK/RESET key is ignored the first four times, but kills Kermit the 5th time). In any case:
When HP-UX is on the remote end of the connection, it is essential that HP-UX C-Kermit be configured for Xon/Xoff flow control (this is the default, but in case you change it and then experience file-transfer failures, this is a likely reason).
This section applies mainly to old (pre-10.20) HP-UX version on old, slow, and/or memory-constrained hardware.
During the C-Kermit 6.0 Beta cycle, something happened to ckcpro.w (or, more precisely, the ckcpro.c file that is generated from it) which causes HP optimizing compilers under HP-UX versions 7.0 and 8.0 (apparently on all platforms) as well as under HP-UX 9.0 on Motorola platforms only, to blow up. In versions 7.0 and 8.0 the problem has spread to other modules.
The symptoms vary from the system grinding to a halt, to the compiler crashing, to the compilation of the ckcpro.c module taking very long periods of time, like 9 hours. This problem is handled by compiling the modules that tickle it without optimization; the new C-Kermit makefile takes care of this, and shows how to do it in case the same thing begins happening with other modules.
On HP-UX 9.0, a kernel parameter, maxdsiz (maximum process data segment size), seems to be important. On Motorola systems, it is 16MB by default, whereas on RISC systems the default is much bigger. Increasing maxdsiz to about 80MB seems to make the problem go away, but only if the system also has a lot of physical memory -- otherwise it swaps itself to death.
The optimizing compiler might complain about "some optimizations skipped" on certain modules, due to lack of space available to the optimizer. You can increase the space (the incantation depends on the particular compiler version -- see the makefile), but doing so tends to make the compilations take a much longer time. For example, the "hpux0100o+" makefile target adds the "+Onolimit" compiler flag, and about an hour to the compile time on an HP-9000/730. But it *does* produce an executable that is about 10K smaller :-)
In the makefile, all HP-UX entries automatically skip optimization of problematic modules.
Telnet connections into HP-UX versions up to and including 11.11 (and possibly 11.20) tend not to lend themselves to file transfer due to limitations, restrictions, and/or bugs in the HP-UX Telnet server and/or pseudoterminal (pty) driver.
In C-Kermit 6.0 (1996) an unexpected slowness was noted when transferring files over local Ethernet connections when an HP-UX system (9.05 or 10.00) was on the remote end. The following experiment was conducted to determine the cause. C-Kermit 6.0 was used; the situation is slightly better using C-Kermit 7.0's streaming feature and HP-UX 10.20 on the far end.
The systems were HP-UX 10.00 (on 715/33) and SunOS 4.1.3 (on Sparc-20), both on the same local 10Mbps Ethernet, packet length 4096, parity none, control prefixing "cautious", using only local disks on each machine -- no NFS. In the C-Kermit 6.0 (ACK/NAK) case, the window size was 20; in the streaming case there is no window size (i.e. it is infinite). The test file was C-Kermit executable, transferred in binary mode. Conditions were relatively poor: the Sun and the local net heavily loaded; the HP system is old, slow, and memory-constrained.
C-Kermit 6.0... C-Kermit 7.0...
Local Remote ACK/NAK........ Streaming......
Client Server Send Receive Send Receive
Sun HP 36 18 64 18
HP HP 25 15 37 16
HP Sun 77 83 118 92
Sun Sun 60 60 153 158
So whenever HP is the remote we have poor performance. Why?
BUT... If I start C-Kermit as a TCP server:
set host * 3000 server
and then from the client "set host blah 3000", I get:
C-Kermit 6.0... C-Kermit 7.0...
Local Remote ACK/NAK........ Streaming......
Client Server Send Receive Send Receive
Sun HP 77 67 106 139
HP HP 50 50 64 62
HP Sun 57 85 155 105
Sun Sun 57 50 321 314
Therefore the HP-UX telnet server or pty driver seems to be adding more overhead than the SunOS one, and most others. When going through this type of connection (a remote telnet server) there is little Kermit can do improve matters, since the telnet server and pty driver are between the two Kermits, and neither Kermit program can have any influence over them (except putting the Telnet connection in binary mode, but that doesn't help).
(The numbers for the HP-HP transfers are lower than the others since both Kermit processes are running on the same slow 33MHz CPU.)
Matters seem to have deteriorated in HP-UX 11. Now file transfers over Telnet connections fail completely, rather than just being slow. In the following trial, a Telnet connection was made from Kermit 95 to HP-UX 11.11 on an HP-9000/785/B2000 over local 10Mbps Ethernet running C-Kermit 8.00 in server mode (under the HP-UX Telnet server):
Text........ Binary......
Stream Pktlen GET SEND GET SEND
On 4000 Fail Fail Fail Fail
Off 4000 Fail Fail Fail Fail
Off 2000 OK Fail OK Fail
On 2000 OK Fail OK Fail
On 3000 Fail Fail Fail Fail
On 2500 Fail Fail Fail Fail
On 2047 OK Fail OK Fail
On 2045 OK Fail OK Fail
Off 500 OK OK OK OK
On 500 OK Fail OK Fail
On 240 OK Fail OK Fail
As you can see, downloads are problematic unless the receiver's Kermit packet length is 2045 or less, but uploads work only with streaming disabled and the packet length restricted to 500. To force file transfers to work on this connection, the desktop Kermit must be told to:
set streaming off set receive packet-length 2000 set send packet-length 500
However, if a connection is made between the same two programs on the same two computers over the same network, but this time a direct socket-to-socket connection bypassing the HP-UX Telnet server and pty driver (tell HP-UX C-Kermit to "set host /server * 3000 /raw"; tell desktop client program to "set host blah 3000 /raw"), everything works perfectly with the default Kermit settings (streaming, 4K packets, liberal control-character unprefixing, 8-bit transparency, etc):
Text........ Binary......
Stream Pktlen GET SEND GET SEND
On 4000 OK OK OK OK
And in this case, transfer rates were approximately 900,000 cps. To verify that the behavior reported here is not caused by the new Kermit release, the same experiment was performed on a Telnet connection from the same PC over the same network to the old 715/33 running HP-UX 10.20 and C-Kermit 8.00. Text and binary uploads and downloads worked perfectly (albeit slowly) with all the default settings -- streaming, 4K packets, etc.
HP workstations do not come with dialout devices configured; you have to do it yourself (as root). First look in /dev to see what's there; for example in HP-UX 10.00 or later:
ls -l /dev/cua* ls -l /dev/tty*
If you find a tty0p0 device but no cua0p0, you'll need to creat one if you want to dial out; the tty0p0 does not work for dialing out. It's easy: start SAM; in the main Sam window, double-click on Peripheral Device, then in the Peripheral Devices window, double-click on Terminals and Modems. In the Terminals and Modems dialog, click on Actions, then choose "Add modem" and fill in the blanks. For example: Port number 0, speed 57600 (higher speeds tend not to work reliably), "Use device for calling out", do NOT "Receive incoming calls" (unless you know what you are doing), leave "CCITT modem" unchecked unless you really have one, and do select "Use hardware flow control (RTS/CTS)". Then click OK. This creates cua0p0 as well as cul0p0 and ttyd0p0
If the following sequence:
set line /dev/cua0p0 ; or other device set speed 115200 ; or other normal speed
produces the message "?Unsupported line speed". This means either that the port is not configured for dialout (go into SAM as described above and make sure "Use device for calling out" is selected), or else that speed you have given (such as 460800) is supported by the operating system but not by the physical device (in which case, use a lower speed like 57600).
In HP-UX 9.0, serial device names began to change. The older names looked like "/dev/cua00", "/dev/tty01", etc (sometimes with only one digit). The newer names have two digits with the letter "p" in between. HP-UX 8.xx and earlier have the older form, HP-UX 10.00 and later have the newer form. HP-UX 9.xx has the newer form on Series 800 machines, and the older form on other hardware models. The situation is summarized in the following table (the Convio 10.0 column applies to HP-UX 10 and 11).
Converged HP-UX Serial I/O Filenames : TTY Mux Naming --------------------------------------------------------------------- General meaning Old Form S800 9.0 Convio 10.0 --------------------------------------------------------------------- tty* hardwired ports tty<YY> tty<X>p<Y> tty<D>p<p> diag:mux<X> diag:mux<D> --------------------------------------------------------------------- ttyd* dial-in modems ttyd<YY> ttyd<X>p<Y> ttyd<D>p<p> diag:ttyd<X>p<Y> diag:ttyd<D>p<p> --------------------------------------------------------------------- cua* auto-dial out cua<YY> cua<X>p<Y> cua<D>p<p> diag:cua<X>p<Y> --------------------------------------------------------------------- cul* dial-out cul<YY> cul<X>p<Y> cul<D>p<p> diag:cul<X>p<Y> --------------------------------------------------------------------- <X>= LU (Logical Unit) <D>= Devspec (decimal card instance) <Y> or <YY> = Port <p>= Port
For dialing out, you should use the cua or cul devices. When C-Kermit's CARRIER setting is AUTO or ON, C-Kermit should pop back to its prompt automatically if the carrier signal drops, e.g. when you log out from the remote computer or service. If you use the tty<D>p<d> (e.g. tty0p0) device, the carrier signal should be ignored. The tty<D>p<d> device should be used for direct connections where the carrier signal does not follow RS-232 conventions (use the cul device for hardwired connections through a true null modem). Do not use the ttyd<D>p<d> device for dialing out.
Kermit's access to serial devices is controlled by "UUCP lockfiles", which are intended to prevent different users using different software programs (Kermit, cu, etc, and UUCP itself) from accessing the same serial device at the same time. When a device is in use by a particular user, a file with a special name is created in:
/var/spool/locks (HP-UX 10.00 and later) /usr/spool/uucp (HP-UX 9.xx and earlier)
The file's name indicates the device that is in use, and its contents indicates the process ID (pid) of the process that is using the device. Since serial devices and the locks directory are not both publicly readable and writable, Kermit and other communication software must be installed setuid to the owner (bin) of the serial device and setgid to the group (daemon) of the /var/spool/locks directory. Kermit's setuid and setgid privileges are enabled only when opening the device and accessing the lockfiles.
Let's say "unit" means a string of decimal digits (the interface instance number) followed (in HP-UX 10.00 and later) by the letter "p" (lowercase), followed by another string of decimal digits (the port number on the interface), e.g.:
"0p0", "0p1", "1p0", etc (HP-UX 10.00 and later) "0p0", "0p1", "1p0", etc (HP-UX 9.xx on Series 800) "00", "01", "10", "0", etc (HP-UX 9.xx not on Series 800) "00", "01", "10", "0", etc (HP-UX 8.xx and earlier)
Then a normal serial device (driver) name consists of a prefix ("tty", "ttyd", "cua", "cul", or possibly "cuad" or "culd") followed by a unit, e.g. "cua0p0". Kermit's treatment of UUCP lockfiles is as close as possible to that of the HP-UX "cu" program. Here is a table of the lockfiles that Kermit creates for unit 0p0:
Selection Lockfile 1 Lockfile 2 /dev/tty0p0 LCK..tty0p0 (none) * /dev/ttyd0p0 LCK..ttyd0p0 (none) /dev/cua0p0 LCK..cua0p0 LCK..ttyd0p0 /dev/cul0p0 LCK..cul0p0 LCK..ttyd0p0 /dev/cuad0p0 LCK..cuad0p0 LCK..ttyd0p0 /dev/culd0p0 LCK..culd0p0 LCK..ttyd0p0 <other> LCK..<other> (none)
(* = Dialin device, should not be used.)
In other words, if the device name begins with "cu", a second lockfile for the "ttyd" device, same unit, is created, which should prevent dialin access on that device.
The <other> case allows for symbolic links, etc, but of course it is not foolproof since we have no way of telling which device is really being used.
When C-Kermit tries to open a dialout device whose name ends with a "unit", it searches the lockfile directory for all possible names for the same unit. For example, if user selects /dev/cul2p3, Kermit looks for lockfiles named:
LCK..tty2p3 LCK..ttyd2p3 LCK..cua2p3 LCK..cul2p3 LCK..cuad2p3 LCK..culd2p3
If any of these files are found, Kermit opens them to find out the ID (pid) of the process that created them; if the pid is still valid, the process is still active, and so the SET LINE command fails and the user is informed of the pid so s/he can use "ps" to find out who is using the device.
If the pid is not valid, the file is deleted. If all such files (i.e. with same "unit" designation) are successfully removed, then the SET LINE command succeeds; up to six messages are printed telling the user which "stale lockfiles" are being removed.
When the "set line" command succeeds in HP-UX 10.00 and later, C-Kermit also creates a Unix System V R4 "advisory lock" as a further precaution (but not guarantee) against any other process obtaining access to the device while you are using it.
If the selected device was in use by "cu", Kermit can't open it, because "cu" has changed its ownership, so we never get as far as looking at the lockfiles. In the normal case, we can't even look at the device to see who the owner is because it is visible only to its (present) owner. In this case, Kermit says (for example):
/dev/cua0p0: Permission denied
When Kermit releases a device it has successfully opened, it removes all the lockfiles that it created. This also happens whenever Kermit exits "under its own power".
If Kermit is killed with a device open, the lockfile(s) are left behind. The next Kermit program that tries to assign the device, under any of its various names, will automatically clean up the stale lockfiles because the pids they contain are invalid. The behavior of cu and other communication programs under these conditions should be the same.
Here, by the way, is a summary of the differences between the HP-UX port driver types from John Pezzano of HP:
There are three types of device files for each port.The ttydXXX device file is designed to work as follows:
- The process that opens it does NOT get control of the port until CD is asserted. This was intentional (over 15 years ago) to allow getty to open the port but not control it until someone called in. If a process wants to use the direct or callout device files (ttyXXX and culXXX respectively), they will then get control and getty would be blocked. This eliminated the need to use uugetty (and its inherent problems with lock files) for modems. You can see this demonstrated by the fact that "ps -ef" shows a ? in the tty column for the getty process as getty does not have the port yet.
- Once CD is asserted, the port is controlled by getty (or the process handling an incoming call) if there was no process using the port. The ? in the "ps" command now shows the port. At this point, the port accepts data.
Therefore you should use either the callout culXXX device file (immediate control but no data until CD is asserted) or the direct device file ttyXXX which gives immediate control and immediate data and which ignores by default modem control signals.
The ttydXXX device should be used only for callin and my recommendation is to use it only for getty and uugetty.
3.2.4.1. HP-UX 11 3.2.4.2. HP-UX 10 3.2.4.3. HP-UX 9 3.2.4.4. HP-UX 8 3.2.4.5. HP-UX 7 and Earlier
As noted in Section 3.2.2, the HP-UX 11 Telnet server and/or pseudoterminal driver are a serious impediment to file transfer over Telnet connections into HP-UX. If you have a Telnet connection into HP-UX 11, tell your desktop Kermit program to:
set streaming off set receive packet-length 2000 set send packet-length 500
File transfer speeds over connections from HP-UX 11 (dialed or Telnet) are not impeded whatsoever, and can go at whatever speed is allowed by the connection and the Kermit partner on the far end.
PA-RISC binaries for HP-UX 10.20 or later should run on any PA-RISC system, S700 or S800, as long as the binary was not built under a later HP-UX version than the host operating system. HP-UX 11.00 and 11.11 are only for PA-RISC systems. HP-UX 11.20 is only for IA64 (subsequent HP-UX releases will be for both PA-RISC and IA64). To check binary compatibility, the following C-Kermit 8.0 binaries were run successfully on an HP-9000/785 with HP-UX 11.11:
Beginning in HP-UX 10.10, libcurses is linked to libxcurses, the new UNIX95 (X/Open) version of curses, which has some serious bugs; some routines, when called, would hang and never return, some would dump core. Evidently libxcurses contains a select() routine, and whenever C-Kermit calls what it thinks is the regular (sockets) select(), it gets the curses one, causing a segmentation fault. There is a patch for this from HP, PHCO_8086, "s700_800 10.10 libcurses patch", "shared lib curses program hangs on 10.10", "10.10 enhanced X/Open curses core dumps due to using wrong select call", 96/08/02 (you can tell if the patch is installed with "what /usr/lib/libxcurses.1"; the unpatched version is 76.20, the patched one is 76.20.1.2). It has been verified that C-Kermit works OK with the patched library, but results are not definite for HP-UX 10.20 or higher.
To ensure that C-Kermit works even on non-patched HP-UX 10.10 systems, separate makefile entries are provided for HP-UX 10.00/10.01, 10.10, 10.20, etc, in which the entries for 10.10 and above link with libHcurses, which is "HP curses", the one that was used in 10.00/10.01. HP-UX 11.20 and later, however, link with libcurses, as libHcurses disappeared in 11.20.
HP-UX 9.00 and 9.01 need patch PHNE_10572 (note: this replaces PHNE_3641) for hptt0.o, asio0.o, and ttycomn.o in libhp-ux.a. Contact Hewlett Packard if you need this patch. Without it, the dialout device (tty) will be hung after first use; subsequent attempts to use will return an error like "device busy". (There are also equivalent patches for s700 9.03 9.05 9.07 (PHNE_10573) and s800 9.00 9.04 (PHNE_10416).
When C-Kermit is in server mode, it might have trouble executing REMOTE HOST commands. This problem happens under HP-UX 9.00 (Motorola) and HP-UX 9.01 (RISC) IF the C-Shell is the login shell AND with the C-Shell Revision 70.15. Best thing is to install HP's Patch PHCO_4919 for Series 300/400 and PHCO_5015 for the Series 700/800. PHCO_5015 is called "s700_800 9.X cumulative csh(1) patch with memory leak fix" which works for HP-UX 9.00, 9.01, 9.03, 9.04, 9.05 and 9.07. At least you need C-Shell Revision 72.12!
C-Kermit works fine -- including its curses-based file-transfer display -- on the console terminal, in a remote session (e.g. when logged in to the HP 9000 on a terminal port or when telnetted or rlogin'd), and in an HP-VUE hpterm window or an xterm window.
To make C-Kermit work on HP-UX 8.05 on a model 720, obtain and install HP-UX patch PHNE_0899. This patch deals with a lot of driver issues, particularly related to communication at higher speeds.
One user reports:
On HP-UX 8 DON'T install 'tty patch' PHKL_4656, install PHKL_3047 instead! Yesterday I tried this latest tty patch PHKL_4656 and had terrible problems. This patch should fix RTS/CTS problems. With text transver all looks nice. But when I switched over to binary files the serial interface returned only rubish to C-Kermit. All sorts of protocol, CRC and packed errors I had. After several tests and after uninstalling that patch, all transvers worked fine. MB's of data without any errors. So keep your fingers away from that patch. If anybody needs the PHKL_3047 patch I have it here. It is no longer availabel from HP's patch base.
When transferring files into HP-UX 5 or 6 over a Telnet connection, you must not use streaming, and you must not use a packet length greater than 512. However, you can use streaming and longer packets when sending files from HP-UX on a Telnet connection. In C-Kermit 8.0, the default receive packet length for HP-UX 5 and 6 was changed to 500 (but you can still increase it with SET RECEIVE PACKET-LENGTH if you wish, e.g. for non-Telnet connections). Disable streaming with SET STREAMING OFF.
The HP-UX 5.00 version of C-Kermit does not include the fullscreen file-transfer because of problems with the curses library.
If HP-UX 5.21 with Wollongong TCP/IP is on the remote end of a Telnet connection, streaming transfers to HP-UX invariably fail. Workaround: SET STREAMING OFF. Packets longer than about 1000 should not be used. Transfers from these systems, however, can use streaming and/or longer packets.
Reportedly, "[there is] a bug in C-Kermit using HP-UX version 5.21 on the HP-9000 series 500 computers. It only occurs when the controlling terminal is using an HP-27140 six-port modem mux. The problem is not present if the controlling terminal is logged into an HP-27130 eight-port mux. The symptom is that just after dialing successfully and connecting Kermit locks up and the port is unusable until both forks of Kermit and the login shell are killed." (This report predates C-Kermit 6.0 and might no longer apply.)
Although C-Kermit presently does not include built-in support for HP-UX X.25 (as it does for the Sun and IBM X.25 products), it can still be used to make X.25 connections as follows: start Kermit and then telnet to localhost. After logging back in, start padem as you would normally do to connect over X.25. Padem acts as a pipe between Kermit and X.25. In C-Kermit 7.0, you might also be able to avoid the "telnet localhost" step by using:
C-Kermit> pty padem address
This works if padem uses standard i/o (who knows?).
SECTION CONTENTS
3.3.1. Problems Building C-Kermit for Linux 3.3.2. Problems with Serial Devices in Linux 3.3.3. Terminal Emulation in Linux 3.3.4. Dates and Times 3.3.5. Startup Errors 3.3.6. The Fullscreen File Transfer Display
REFERENCES
For further information, read the comp.os.linux.misc, comp.os.linux.answers, and other Linux-oriented newsgroups, and see:
Also see general comments on PC-based Unixes in Section 3.0.
What Linux version is it? -- "uname -a" supplies only kernel information, but these days it's the distribution that matters: Red Hat 7.3, Debian 2.2, Slackware 8.0, etc. Unfortunately there's no consistent way to get the distribution version. Usually it's in a distribution-specific file:
| Red Hat: | /etc/issue or /etc/redhat-release |
| Debian: | /etc/debian_version |
| Slackware: | /etc/slackware-version (at least in later versions) |
Did you know: DECnet is available for Linux? See:
http://linux.dreamtime.org/decnet/
(But there is no support for it in C-Kermit -- anybody interested in adding it, please let us know).
Before proceeding, let's handle the two most frequently asked question in the Linux newsgroups:
Modern Linux distributions like Red Hat give you a choice at installation whether to include "developer tools". Obviously, you can't build C-Kermit or any other C program from source code if you have not installed the developer tools. But to confuse matters, you might also have to choose (separately) to install the "curses" or "ncurses" terminal control library; thus it is possible to install the C compiler and linker, but omit the (n)curses library and headers. If curses is not installed, you will not be able to build a version of C-Kermit that supports the fullscreen file-transfer display, in which case you'll need to use the "linuxnc" makefile target (nc = No Curses) or else install ncurses before building.
Be sure to read the comments in the "linux:" makefile entry. There are all sorts of confusing issues caused by the many and varied Linux distributions. Some of the worst involve the curses library and header files: where are they, what are they called, which ones are they really? Other vexing questions involve libc5 vs libc6 vs glibc vs glibc2 (C libraries), gcc vs egcs vs lcc (compilers), plus using or avoiding features that were added in a certain version of Linux or a library or a distribution, and are not available in others.
Linux C-Kermit, like all other Unix C-Kermit versions, was built traditionally with curses.h and the curses library. However, this library was evidently so buggy (users reported that, after doing a file transfer using the fullscreen display, "screen scrolling locks up" and the cursor "is stuck on the bottom of the screen", etc), that a new curses library, called ncurses, was developed to replace it. The Linux version of C-Kermit, as of version 6.1, uses ncurses rather than curses.
After modern practice, ncurses is dynamically linked, rather than linked into the executable. This means a certain relationship must obtain between the version number referenced in the executable and the version number of the library. But there are evidently several different numbering systems for libncurses.so -- e.g. 1.9.9e is another "name" for 3.0 -- but the program loader doesn't know that and so won't run the program. Also the library and/or terminfo database might be in a different place on the target system (e.g. /usr/share/terminfo) than it was on the build system (e.g. /usr/lib/terminfo). Solution: Create the appropriate symbolic links and/or rebuild C-Kermit yourself from source code, and if you have additional trouble, come back and read the rest of this section.
Of course static linking is also a possibility, but this makes the executable MUCH bigger and introduces new problems of its own.
From the March 1999 Kermit newsgroup traffic:
: When I start Kermit (under Redhat Linux 5.2), it complains about not : being able to recognise my terminal type - I've tried all the obvious : terminal types - which ones can I use? Or can I get it to recognise : xterm? : Assuming that you can use full screen programs, this looks identical to the problem introduced by RedHat with 5.1. They moved the curses library, and didn't [ leave a link from the old location to the new one ]: To fix: cd /usr/share; ln -s terminfo ../lib
The termcap library is no longer referenced in the Linux target in the makefile, since its functions are supposedly incorporated into the ncurses and curses libraries. However, should any termcap-related entry points come up undefined at link time (_tgetent, _tgoto, _tputs, etc), it might be necessary to add -ltermcap back to the "LIBS=" clause. But then you might find that the termcap library is not in /usr/lib after all, but has been moved to /usr/lib/termcap/, in which case you'll need to make a symlink, or do something like:
"LIBS = -L/usr/lib/curses -lcurses -L/usr/lib/termcap -ltermcap"
Different UUCP lockfile conventions are used by different Linux versions and/or distributions. In C-Kermit 6.0 and later, "make linux" uses /var/lock/LCK..name, decimal ASCII 10-byte PID string with leading spaces because -DLINUXFSSTND ("Linux File System Standard") is included in the compilation CFLAGS. If you remove this definition, C-Kermit will use the earlier arrangement of integer PID, /usr/spool/uucp/LCK..name. The leading spaces are required by FSSTND 1.2, but FSSTND 1.0 required leading zeros; to get the leading zeros, also include -DFSSTND10. Use whichever option agrees with your uucp, cu, tip, etc, programs.
One user reported problems building C-Kermit under Linux 2.0.30/Slackware 96, errors like:
/usr/include/linux/socket.h:77: warning: `PF_AAL5' redefined /usr/local/include/socketbits.h:68: warning: this is the location of the previous definition ckutio.c:4679: `TIOCGSERIAL' undeclared (first use this function) ckutio.c:4685: `TIOCSSERIAL' undeclared (first use this function) ckutio.c:6092: warning: passing arg 3 of `select' from incompatible pointer type
etc etc. Diagnosis: These were caused by installing some other package, which created files in /usr/local/include. Cure:
rm -rf /usr/local/include
start over.
Reportedly, building C-Kermit 6.0 on Linux 1.1.33 and 1.1.34 gets fatal compilation errors due to inconsistencies in the Linux header files. Linux kernel versions prior to 1.1.33 and later than 1.1.34 should be OK. (Also, C-Kermit 6.1 and later should be OK since we no longer include kernel header files.)
Reportedly there is a bug in gcc 2.5.8 with signed to unsigned compares that can wreak havoc when Kermit (or most any other program) is compiled with this version of gcc; reportedly this can be worked around, at least in part, by adding "-fno-unroll-loops" to the gcc compilation options. (This problem is evidently fixed in more recent gcc releases.)
Reportedly, if you have the iBCS2 (Intel Binary Compatibility Standard 2) module installed, you can also run SCO Xenix and Unix binaries under Linux, including the SCO C-Kermit binaries, shareable libraries and all. (iCBS2 is available via anonymous ftp from tsx-11.mit.edu, along with an SCO libc_s compatibility module for Linux).
There is evidently a minor problem with GCC (version unknown) on (64-bit) Alpha platforms, in which it complains:
warning: cast to pointer from integer of different size
whenever it encounters a legitimate trinary expression like:
integer ? "string1" : "string2"
(The "integer" can also be an integer-valued expression.) These warnings appear to be harmless.
Also see: "man setserial", "man irqtune".
And: Sections 3.0, 6, 7, and 8 of this document.
NOTE: Red Hat Linux 7.2 and later include a new API that allows serial-port arbitration by non-setuid/gid programs. This API has not yet been added to C-Kermit. If C-Kermit is to be used for dialing out on Red Hat 7.2 or later, it must still be installed as described in in Sections 10 and 11 of the Installation Instructions.
Don't expect it to be easy. Queries like the following are posted to the Linux newsgroups almost daily:
Problem of a major kind with my Compaq Presario 1805 in the sense that the pnpdump doesn't find the modem and the configuration tells me that the modem is busy when I set everything by hand!I have <some recent SuSE distribution>, kernel 2.0.35. Using the Compaq tells me that the modem (which is internal) is on COM2, with the usual IRQ and port numbers. Running various Windows diagnostics show me AT-style commands exchanged so I have no reason to beleive that it is a Winmodem. Also, the diagnostics under Win98 tell me that I am talking to an NS 16550AN.
[Editor's note: This does not necessarily mean it isn't a Winmodem.]
Under Linux, no joy trying to talk to the modem on /dev/cua1 whether via minicom, kppp, or chat; kppp at least tells me that tcgetattr() failed.Usage of setserial:
setserial /dev/cua1 port 0x2F8 irq 3 autoconfig setserial -g /dev/cua1tells me that the uart is 'unknown'. I have tried setting the UART manullay via. setserial to 16550A, 16550, and the other one (8550?) (I didn't try 16540). None of these manual settings resulted in any success.
A look at past articles leads me to investigate PNP issues by calling pnpdump but pnpdump returns "no boards found". I have looked around on my BIOS (Phoenix) and there is not much evidence of it being PNP aware. However for what it calls "Serial port A", it offers a choice of Auto, Disabled or Manual settings (currently set to Auto), but using the BIOS interface I tried to change to 'manual' and saw the default settings offered to be were 0x3F8 and IRQ 4 (COM1). The BIOS menus did not give me any chance to configure COM2 or any "modem". I ended up not saving any BIOS changes in the course of my investigations.
You can also find out a fair amount about your PC's hardware configuration in the text files in /proc, e.g.:
-r--r--r-- 1 root 0 Sep 4 14:00 /proc/devices -r--r--r-- 1 root 0 Sep 4 14:00 /proc/interrupts -r--r--r-- 1 root 0 Sep 4 14:00 /proc/ioports -r--r--r-- 1 root 0 Sep 4 14:00 /proc/pci
From the directory listing they look like empty files, but in fact they are text files that you "cat":
$ cat /proc/pci
Bus 0, device 14, function 0:
Serial controller: US Robotics/3Com 56K FaxModem Model 5610 (rev 1).
IRQ 10.
I/O at 0x1050 [0x1057].
$ setserial -g /dev/ttyS4
/dev/ttyS4, UART: 16550A, Port: 0x1050, IRQ: 10
$ cat /proc/ioports
1050-1057 : US Robotics/3Com 56K FaxModem Model 5610
1050-1057 : serial(auto)
$ cat /proc/interrupts
CPU0
0: 7037515 XT-PIC timer
1: 2 XT-PIC keyboard
2: 0 XT-PIC cascade
4: 0 XT-PIC serial
8: 1 XT-PIC rtc
9: 209811 XT-PIC usb-uhci, eth0
14: 282015 XT-PIC ide0
15: 6 XT-PIC ide1
Watch out for PCI, PCMCIA and Plug-n-Play devices, Winmodems, and the like (see cautions in Section 3.0 Linux supports Plug-n-Play devices to some degree via the isapnp and pnpdump programs; read the man pages for them. (If you don't have them, look on your installation CD for isapnptool or download it from sunsite or a sunsite mirror or other politically correct location).
PCI modems do not use standard COM port addresses. The I/O address and IRQ are assigned by the BIOS. All you need to do to get one working, find out the I/O address and interrupt number with (as root) "lspci -v | more" and then give the resulting address and interrupt number to setserial.
Even when you have a real serial port, always be wary of interrupt conflicts and similar PC hardware configuration issues: a PC is not a real computer like other Unix workstations -- it is generally pieced together from whatever random components were the best bargain on the commodity market the week it was built. Once it's assembled and boxed, not even the manufacturer will remember what it's made of or how it was put together because they've moved on to a new model. Their job is to get it (barely) working with Windows; for Linux and other OS's you are on your own.
"set line /dev/modem" or "set line /dev/ttyS2", etc, results in an error, "/dev/modem is not a tty". Cause unknown, but obviously a driver issue, not a Kermit one (Kermit uses "isatty()" to check that the device is a tty, so it knows it will be able to issue all the tty-related ioctl's on it, like setting the speed & flow control). Try a different name (i.e. driver) for the same port, e.g. "set line /dev/cua2" or whatever.
To find what serial ports were registered at the most recent system boot, type (as root): "grep tty /var/log/dmesg".
"set modem type xxx" (where xxx is the name of a modem) followed by "set line /dev/modem" or "set line /dev/ttyS2", etc, hangs (but can be interrupted with Ctrl-C). Experimentation shows that if the modem is configured to always assert carrier (&C0) the same command does not hang. Again, a driver issue. Use /dev/cua2 (or whatever) instead. (Or not -- hopefully none of these symptoms occurs in C-Kermit 7.0 or later.)
"set line /dev/cua0" reports "Device is busy", but "set line /dev/ttyS0" works OK.
In short: If the cua device doesn't work, try the corresponding ttyS device. If the ttyS device doesn't work, try the corresponding cua device -- but note that Linux developers do not recommend this, and are phasing out the cua devices. From /usr/doc/faq/howto/Serial-HOWTO:
It was discovered during development of C-Kermit 7.0 that rebuilding C-Kermit with -DNOCOTFMC (No Close/Open To Force Mode Change) made the aforementioned problem with /dev/ttyS0 go away. It is not yet clear, however, what its affect might be on the /dev/cua* devices. As of 19 March 1998, this option has been added to the CFLAGS in the makefile entries for Linux ("make linux").
Note that the cua device is now "deprecated", and new editions of Linux will phase (have phased) it out in favor of the ttyS device. See (if it's still there):
http://linuxwww.db.erau.edu/mail_archives/linux-kernel/Mar_98/1441.html
(no, of course it isn't; you'll have to use your imagination). One user reported that C-Kermit 7.0, when built with egcs 1.1.2 and run on Linux 2.2.6 with glibc 2.1 (hardware unknown but probably a PC) dumps core when given a "set line /dev/ttyS1" command. When rebuilt with gcc, it works fine.
All versions of Linux seem to have the following deficiency: When a modem call is hung up and CD drops, Kermit can no longer read the modem signals; SHOW COMMUNICATIONS says "Modem signals not available". The TIOCMGET ioctl() returns -1 with errno 5 ("I/O Error").
The Linux version of POSIX tcsendbreak(), which is used by C-Kermit to send regular (275msec) and long (1.5sec) BREAK signals, appears to ignore its argument (despite its description in the man page and info topic), and always sends a regular 275msec BREAK. This has been observed in Linux versions ranging from Debian 2.1 to Red Hat 7.1.
C-Kermit is not a terminal emulator. For a brief explanation of why not, see Section 3.0.5. For a fuller explanation, ClICK HERE.
In Unix, terminal emulation is supplied by the Window in which you run Kermit: the regular console screen, which provides Linux Console "emulation" via the "console" termcap entry, or under X-Windows in an xterm window, which gives VTxxx emulation. An xterm that includes color ANSI and VT220 emulation is available with Xfree86:
http://dickey.his.com/xterm/xterm.html
Before starting C-Kermit in an xterm window, you might need to tell the xterm window's shell to "stty sane".
To set up your PC console keyboard to send VT220 key sequences when using C-Kermit as your communications program in an X terminal window (if it doesn't already), create a file somewhere (e.g. in /root/) called .xmodmaprc, containing something like the following:
keycode 77 = KP_F1 ! Num Lock => DEC Gold (PF1) keycode 112 = KP_F2 ! Keypad / => DEC PF1 keycode 63 = KP_F3 ! Keypad * => DEC PF3 keycode 82 = KP_F4 ! Keypad - => DEC PF4 keycode 111 = Help ! Print Screen => DEC Help keycode 78 = F16 ! Scroll Lock => DEC Do keycode 110 = F16 ! Pause => DEC Do keycode 106 = Find ! Insert => DEC Find keycode 97 = Insert ! Home => DEC Insert keycode 99 = 0x1000ff00 ! Page Up => DEC Remove keycode 107 = Select ! Delete => DEC Select keycode 103 = Page_Up ! End => DEC Prev Screen keycode 22 = Delete ! Backspace sends Delete (127)
Then put "xmodmap filename" in your .xinitrc file (in your login directory), e.g.
xmodmap /root/.xmodmaprc
Of course you can move things around. Use the xev program to find out key codes.
Console-mode keys are mapped separately using loadkeys, and different keycodes are used. Find out what they are with showkey.
For a much more complete VT220/320 key mapping for Xfree86 xterm, CLICK HERE.
If C-Kermit's date-time (e.g. as shown by its DATE command) differs from the system's date and time:
C-Kermit should work on all versions of Linux current through December 2001, provided it was built on the same version you have, with the same libraries and header files (just get the source code and "make linux"). Binaries tend not to travel well from one Linux machine to another, due to their many differences. There is no guarantee that a particular C-Kermit binary will not stop working at a later date, since Linux tends to change out from under its applications. If that happens, rebuild C-Kermit from source. If something goes wrong with the build process, let us know.
Inability to transfer files in Red Hat 7.2: the typical symptom would be if you start Kermit and tell it to RECEIVE, it fails right away with "?/dev/tty: No such device or address" or "?Bad file descriptor". One report says this is because of csh, and if you change your shell to bash or other shell, it doesn't happen. Another report cite bugs in Red Hat 7.2 Telnetd "very seldom (if ever) providing a controlling tty, and lots of other people piled on saying they have the same problem.") A third theory is that this happens only when Linux has been installed without virtual terminal support.
A search of RedHat's errata pages shows a bug advisory (RHBA-2001-153) issued 13 November 2001, but updated 6 December, about this same symptom (but with tcsh and login.) Seems that login was not always assigning a controlling TTY for the session, which would make most use of "/dev/tty" somewhat less than useful.
http://www.redhat.com/support/errata/RHBA-2001-153.html
Quoting: "Due to terminal handling problems in /bin/login, tcsh would not find the controlling terminal correctly, and a shell in single user mode would exhibit strange terminal input characteristics. This update fixes both of these problems."
Since the Red Hat 5.1 release (circa August 1998), there have been numerous reports of prebuilt Linux executables, and particularly the Kermit RPM for Red Hat Linux, not working; either it won't start at all, or it gives error messages about "terminal type unknown" and refuses to initialize its curses support. The following is from the Kermit newsgroup:
From: rchandra@hal9000.buf.servtech.com
Newsgroups: comp.protocols.kermit.misc
Subject: Red Hat Linux/Intel 5.1 and ncurses: suggestions
Date: 22 Aug 1998 15:54:46 GMT
Organization: Verio New York
Keywords: RedHat RPM 5.1Several factors can influence whether "linux" is recognized as a terminal type on many Linux systems.
- Your program, or the libraries it linked with (if statically linked), or the libraries it dynamically links with at runtime, are looking for an entry in /etc/termcap that isn't there. (not likely, but possible... I believe but am not certain that this is a very old practice in very old [n]curses library implementations to use a single file for all terminal descriptions.)
- Your program, or the libraries...are looking for a terminfo file that just plain isn't there. (also not so likely, since many people in other recent message threads said that other programs work OK).
- Your program, or the libraries...are looking for a terminfo file that is stored at a pathname that isn't expected by your program, the libraries--and so on. I forgot if I read this in the errata Web page or where exactly I discovered this (Netscape install? Acrobat install?), but it may just be that one libc (let's say for sake of argument, libc5, but I don't know this to be true) expects your terminfo to be in /usr/share/terminfo, and the other (let's say libc6/glibc) expects /usr/lib/terminfo. I remember that the specific instructions in this bugfix/workaround were to do the following or equivalent:
cd /usr/lib ln -s ../share/terminfo ./terminfoor:
ln -s /usr/share/terminfo /usr/lib/terminfoSo what this says is that the terminfo database/directory structure can be accessed by either path. When something goes to reference /usr/lib/terminfo, the symlink redirects it to essentially /usr/share/terminfo, which is where it really resides on your system. I personally prefer wherever possible to use relative symlinks, because they still hold, more often than break, across mount points, particularly NFS mounts, where the directory structure may be different on the different systems.
Evidently the terminfo file moved between Red Hat 5.0 and 5.1, but Red Hat did not include a link to let applications built prior to 5.1 find it. Users reported that installing the link fixes the problem.
Starting with ncurses versions dated 1998-12-12 (about a year before ncurses 5.0), ncurses sets the terminal for buffered i/o, but unfortunately is not able to restore it upon exit from curses (via endwin()). Thus after a file transfer that uses the fullscreen file transfer display, the terminal no longer echos nor responds immediately to Tab, ?, and other special command characters. The same thing happens on other platforms that use ncurses, e.g. FreeBSD. Workarounds:
In Red Hat 7.1, when using C-Kermit in a Gnome terminal window, it was noticed that when the fullscreen file transfer display exits (via endwin()), the previous (pre-file-transfer-display) screen is restored. Thus you can't look at the completed display to see what happened. This is a evidently a new feature of xterm. I can only speculate that initscreen() and endwin() must send some kind of special escape sequences that command xterm to save and restore the screen. To defeat this effect, tell Linux you have a vt100 or other xterm-compatible terminal that is not actually an xterm, or else tell Kermit to SET TRANSFER DISPLAY to something besides FULLSCREEN.
Run C-Kermit in a Terminal, Stuart, or xterm window, or when logged in remotely through a serial port or TELNET connection. C-Kermit does not work correctly when invoked directly from the NeXTSTEP File Viewer or Dock. This is because the terminal-oriented gtty, stty, & ioctl calls don't work on the little window that NeXTSTEP pops up for non-NeXTSTEP applications like Kermit. CBREAK and No-ECHO settings do not take effect in the command parser -- commands are parsed strictly line at a time. "set line /dev/cua" works. During CONNECT mode, the console stays in cooked mode, so characters are not transmitted until carriage return or linefeed is typed, and you can't escape back. If you want to run Kermit directly from the File Viewer, then launch it from a shell script that puts it in the desired kind of window, something like this (for "Terminal"):
Terminal -Lines 24 -Columns 80 -WinLocX 100 -WinLocY 100 $FONT $FONTSIZE \ -SourceDotLogin -Shell /usr/local/bin/kermit &
C-Kermit does not work correctly on a NeXT with NeXTSTEP 3.0 to which you have established an rlogin connection, due to a bug in NeXTSTEP 3.0, which has been reported to NeXT.
The SET CARRIER command has no effect on the NeXT -- this is a limitation of the NeXTSTEP serial-port device drivers.
Hardware flow control on the NeXT is selected not by "set flow rts/cts" in Kermit (since NeXTSTEP offers no API for this), but rather, by using a specially-named driver for the serial device: /dev/cufa instead /dev/cua; /dev/cufb instead of /dev/cub. This is available only on 68040-based NeXT models (the situation for Intel NeXTSTEP implementations is unknown).
NeXT-built 68030 and 68040 models have different kinds of serial interfaces; the 68030 has a Macintosh-like RS-422 interface, which lacks RTS and CTS signals; the 68040 has an RS-423 (RS-232 compatible) interface, which supports the commonly-used modem signals. WARNING: the connectors look exactly the same, but the pins are used in completely DIFFERENT ways -- different cables are required for the two kinds of interfaces.
IF YOU GET LOTS OF RETRANSMISSIONS during file transfer, even when using a /dev/cuf* device and the modem is correctly configured for RTS/CTS flow control, YOU PROBABLY HAVE THE WRONG KIND OF CABLE.
On the NeXT, Kermit reportedly (by TimeMon) causes the kernel to use a lot of CPU time when using a "set line" connection. That's because there is no DMA channel for the NeXT serial port, so the port must interrupt the kernel for each character in or out.
One user reported trouble running C-Kermit on a NeXT from within NeXT's Subprocess class under NeXTstep 3.0, and/or when rlogin'd from one NeXT to another: Error opening /dev/tty:, congm: No such device or address. Diagnosis: Bug in NeXTSTEP 3.0, cure unknown.
See also: The comp.os.qnx newsgroup.
Support for QNX 4.x was added in C-Kermit 5A(190). This is a full-function implementation, thoroughly tested on QNX 4.21 and later, and verified to work in both 16-bit and 32-bit versions. The 16-bit version was dropped in C-Kermit 7.0 since it can no longer be built successfully (after stripping most most features, I succeeded in getting it to compile and link without complaint, but the executable just beeps when you run it); for 16-bit QNX 4.2x, use C-Kermit 6.0 or earlier, or else G-Kermit.
The 32-bit version (and the 16-bit version prior to C-Kermit 7.0) support most of C-Kermit's advanced features including TCP/IP, high serial speeds, hardware flow-control, modem-signal awareness, curses support, etc.
BUG: In C-Kermit 6.0 on QNX 4.22 and earlier, the fullscreen file transfer display worked fine the first time, but was fractured on subsequent file transfers. Cause and cure unknown. In C-Kermit 7.0 and QNX 4.25, this no longer occurs. It is not known if it would occur in C-Kermit 7.0 or later on earlier QNX versions.
Dialout devices are normally /dev/ser1, /dev/ser2, ..., and can be opened explicitly with SET LINE. Reportedly, "/dev/ser" (no unit number) opens the first available /dev/sern device.
Like all other Unix C-Kermit implementations, QNX C-Kermit does not provide any kind of terminal emulation. Terminal specific functions are provided by your terminal, terminal window (e.g. QNX Terminal or xterm), or emulator.
QNX C-Kermit, as distributed, does not include support for UUCP line-locking; the QNX makefile entries (qnx32 and qnx16) include the -DNOUUCP switch. This is because QNX, as distributed, does not include UUCP, and its own communications software (e.g. qterm) does not use UUCP line locking. If you have a UUCP product installed on your QNX system, remove the -DNOUUCP switch from the makefile entry and rebuild. Then check to see that Kermit's UUCP lockfile conventions are the same as those of your UUCP package; if not, read the UUCP lockfile section of the Installation Instructions and make the necessary changes to the makefile entry (e.g. add -DHDBUUCP).
QNX does, however, allow a program to get the device open count. This can not be a reliable form of locking unless all applications do it, so by default, Kermit uses this information only for printing a warning message such as:
C-Kermit>set line /dev/ser1 WARNING - "/dev/ser1" looks busy...
However, if you want to use it as a lock, you can do so with:
SET QNX-PORT-LOCK { ON, OFF }
This is OFF by default; if you set in ON, C-Kermit will fail to open any dialout device when its open count indicates that another process has it open. SHOW COMM (in QNX only) displays the setting, and if you have a port open, it also shows the open count.
As of C-Kermit 8.0, C-Kermit's "open-count" form of line locking works only in QNX4, not in QNX6 (this might change in a future C-Kermit release).
SECTION CONTENTS
3.6.1. SCO XENIX 3.6.2. SCO UNIX and OSR5 3.6.3. Unixware 3.6.4. Open UNIX 8
REFERENCES
Also see general comments on PC-based Unixes in Section 3.0.
Old Xenix versions... Did you know: Xenix 3.0 is *older* than Xenix 2.0?
In Xenix 2.3.4 and probably other Xenix versions, momentarily dropping DTR to hang up a modem does not work. DTR goes down but does not come up again. Workaround: Use SET MODEM HANGUP-METHOD MODEM-COMMAND. Anybody who would like to fix this is welcome to take a look at tthang() in ckutio.c. Also: modem signals can not be read in Xenix, and the maximum serial speed is 38400.
There is all sorts of confusion among SCO versions, particularly when third- party communications boards and drivers are installed, regarding lockfile naming conventions, as well as basic functionality. As far as lockfiles go, all bets are off if you are using a third-party multiport board. At least you have the source code. Hopefully you also have a C compiler :-)
Xenix 2.3.0 and later claim to support RTSFLOW and CTSFLOW, but this is not modern bidirectional hardware flow control; rather it implements the original RS-232 meanings of these signals for unidirectional half-duplex line access: If both RTSFLOW and CTSFLOW bits are set, Xenix asserts RTS when it wants to send data and waits for CTS assertion before it actually starts sending data (also, reportedly, even this is broken in Xenix 2.3.0 and 2.3.1).
SCO systems tend to use different names (i.e. drivers) for the same device. Typically /dev/tty1a refers to a terminal device that has no modem control; open, read, write, and close operations do not depend on carrier. On the other hand, /dev/tty1A (same name, but with final letter upper case), is the same device with modem control, in which carrier is required (the SET LINE command does not complete until carrier appears, read/write operations fail if there is no carrier, etc).
SCO OpenServer 5.0.5 and earlier do not support the reading of modem signals. Thus "show comm" does not list modem signals, and C-Kermit does not automatically pop back to its prompt when the modem hangs up the connection (drops CD). The ioctl() call for this is simply not implmented, at least not in the standard drivers. OSR5.0.6 attempts to deal with modem signals but fails; however OSR5.0.6a appears to function properly.
Dialing is likely not to work well in SCO OpenServer 5.0.x because many of the serial-port APIs simply do not operate when using the standard drivers. For example, if DTR is dropped by the recommended method (setting speed to 0 for half a seconds, then restoring the speed), DTR and RTS go down but never come back up. When in doubt SET MODEM HANGUP-METHOD MODEM-COMMAND or SET DIAL HANGUP OFF.
On the other hand, certain functions that might not (do not) work right or at all when using SCO drivers (e.g. high serial speeds, hardware flow control, and/or reading of modem signals) might work right when using third-party drivers. (Example: hardware flow control works, reportedly, only on uppercase device like tty1A -- not tty1a -- and only when CLOCAL is clear when using the SCO sio driver, but there are no such restrictions in, e.g., Digiboard drivers).
One user reports that he can't transfer large files with C-Kermit under SCO OSR5.0.0 and 5.0.4 -- after the first 5K, everything falls apart. Same thing without Kermit -- e.g. with ftp over a PPP connection. Later, he said that replacing SCO's SIO driver with FAS, an alternative communications driver, made the problem go away:
ftp://ftp.fu-berlin.de/pub/unix/driver/fas
With regard to bidirectional serial ports on OpenServer 5.0.4, the following advice appeared on an SCO-related newsgroup:
No amount of configuration information is going to help you on 5.0.4 unless it includes the kludge for the primary problem. With almost every modem, the 5.0.4 getty will barf messages and may or may not connect. There are 2 solutions and only one works on 5.0.4. Get the atdialer binary from a 5.0.0 system and substitute it for the native 5.0.4 atdialer. The other solution is to upgrade to 5.0.5. And, most of all, on any OpenServer products, do NOT run the badly broken Modem Manager. Configure the modems in the time honored way that dates back to Xenix.
Use SCO-provided utilities for switching the directionality of a modem line, such as "enable" and "disable" commands. For example, to dial out on tty1a, which is normally set up for logins:
disable tty1a kermit -l /dev/tty1a enable tty1a
If a tty device is listed as an ACU in /usr/lib/uucp/Devices and is enabled, getty resets the ownership and permissions to uucp.uucp and 640 every time the device is released. If you want to use the device only for dialout, and you want to specify other owners or permissions, you should disable it in /usr/lib/uucp/Devices; this will prevent getty from doing things to it. You should also changes the device's file modes in /etc/conf/node.d/sio by changing fields 5-7 for the desired device(s); this determines how the devices are set if you relink the kernel.
One SCO user of C-Kermit 5A(190) reported that only one copy of Kermit can run at a time when a Stallion Technologies multiport boards are installed. Cause, cure, and present status unknown (see Section 14 for more info regarding Stallion).
Prior to SCO OpenServer 5.0.4, the highest serial port speed supported by SCO was 38400. However, in some SCO versions (e.g. OSR5) it is possible to map rarely-used lower speeds (like 600 and 1800) to higher ones like 57600 and 115200. To find out how, go to http://www.sco.com/ and search for "115200". In OSR5.0.4, serial speeds up to 921600 are supported through the POSIX interface; C-Kermit 6.1.193 or later, when built for OSR5.0.4 using /bin/cc (NOT the UDK, which hides the high-speed definitions from CPP), supports these speeds, but you might be able to run this binary on earlier releases to get the high serial speeds, depending on various factors, described by Bela Lubkin of SCO:
Serial speeds under SCO Unix / Open Desktop / OpenServer ======================================================== Third party drivers (intelligent serial boards) may provide any speeds they desire; most support up to 115.2Kbps.
SCO's "sio" driver, which is used to drive standard serial ports with 8250/16450/16550 and similar UARTs, was limited to 38400bps in older releases. Support for rates through 115.2Kbps was added in the following releases:
SCO OpenServer Release 5.0.0 (requires supplement "rs40b")
SCO OpenServer Release 5.0.2 (requires supplement "rs40a" or "rs40b")
SCO OpenServer Release 5.0.4 or later
SCO Internet FastStart Release 1.0.0 or later
SCO supplements are at ftp://ftp.sco.com/; the "rs40" series are under directory /Supplements/internet
Kermit includes the high serial speeds in all OSR5 builds, but that does not necessarily mean they work. For example, on our in-house 5.0.5 system, SET SPEED 57600 or higher seems to succeed (no error occurs) but when we read the speed back the driver says it is 50. Similarly, 76800 becomes 75, and 115200 becomes 110. Testing shows the resulting speed is indeed the low one we read back, not the high one we asked for. Moral: Use speeds higher than 38400 with caution on SCO OSR5.
Reportedly, if you have a script that makes a TCP/IP SET HOST (e.g. Telnet) connection to SCO 3.2v4.2 with TCP/IP 1.2.1, and then does the following:
script $ exit hangup
this causes a pseudoterminal (pty) to be consumed on the SCO system; if you do it enough times, it will run out of ptys. An "exit" command is being sent to the SCO shell, and a HANGUP command is executed locally, so the chances are good that both sides are trying to close the connection at once, perhaps inducing a race condition in which the remote pty is not released. It was speculated that this would be fixed by applying SLS net382e, but it did not. Meanwhile, the workaround is to insert a "pause" between the SCRIPT and HANGUP commands. (The situation with later SCO releases is not known.)
SCO UNIX and OpenServer allow their console and/or terminal drivers to be configured to translate character sets for you. DON'T DO THIS WHEN USING KERMIT! First of all, you don't need it -- Kermit itself already does this for you. And second, it will (a) probably ruin the formatting of your screens (depending on which emulation you are using); and (b) interfere with all sorts of other things -- legibility of non-ASCII text on the terminal screen, file transfer, etc. Use:
mapchan -n
to turn off this feature.
Note that there is a multitude of SCO entries in the makefile, many of them exhibiting an unusually large number of compiler options. Some people actually understand all of this. Reportedly, things are settling down with SCO OpenServer 5.x and Unixware 7 (and Open UNIX 8 and who knows what the next one will be -- Linux probably) -- the SCO UDK compiler is said to generate binaries that will run on either platform, by default, automatically. When using gcc or egcs, on the other hand, differences persist, plus issues regarding the type of binary that is generated (COFF, ELF, etc), and where and how it can run. All of this could stand further clarification by SCO experts.
Unixware changed hands several times before landing at SCO, and so has its own section in this document. (Briefly: AT&T UNIX Systems Laboratories sold the rights to the UNIX name and to System V R4 (or R5?) to Novell; later Novell spun its UNIX division off into a new company called Univel, which eventually was bought by SCO.)
SCO was bought by Caldera in 2000 or 2001 and evolved Unixware 7.1 into Caldera Open UNIX 8.00. It's just like Unixware 7.1 as far as Kermit is concerned (the Unixware 7.1 makefile target works for Open UNIX 8.00, and in fact a Unixware 7.1 Kermit binary built on Unixware 7.1 runs under OU8; a separate OU8 makefile target exists simply to generate an appropriate program startup herald).
SECTION CONTENTS
3.7.1. Serial Port Configuration 3.7.2. Serial Port Problems 3.7.3. SunLink X.25 3.7.4. Sun Workstation Keyboard Mapping 3.7.5. Solaris 2.4 and Earlier
REFERENCES
And about serial communications in particular, see "Celeste's Tutorial on Solaris 2.x Modems and Terminals":
http://www.stokely.com/
In particular:
http://www.stokely.com/unix.sysadm.resources/faqs.sun.html
For PC-based Solaris, also see general comments on PC-based Unixes in Section 3.0. Don't expect Solaris or any other kind of Unix to work right on a PC until you resolve all interrupt conflicts. Don't expect to be able to use COM3 or COM4 (or even COM2) until you have configured their addresses and interrupts.
Your serial port can't be used -- or at least won't work right -- until it is enabled in Solaris. For example, you get a message like "SERIAL: Operation would block" when attempting to dial. This probably indicates that the serial port has not been enabled for use with modems. You'll need to follow the instructions in your system setup or management manual, such as (e.g.) the Desktop SPARC Sun System & Network Manager's Guide, which should contain a section "Setting up Modem Software"; read it and follow the instructions. These might (or might not) include running a program called "eeprom", editing some system configuration file (such as, for example:
/platform/i86pc/kernel/drv/asy.conf
and then doing a configuration reboot, or running some other programs like drvconfig and devlinks. "man eeprom" for details.
Also, on certain Sun models like IPC, the serial port hardware might need to have a jumper changed to make it an RS-232 port rather than RS-423.
eeprom applies only to real serial ports, not to "Spiff" devices (serial port expander), in which case setup with Solaris' admintool is required.
Another command you might need to use is pmadm, e.g.:
pmadm -d -p zsmon -s tty3 pmadm -e -p zsmon -s tty3
You can use the following command to check if a process has the device open:
fuser -f /dev/term/3
In some cases, however (according to Sun support, May 2001) "It is still possible that a zombie process has hold of the port EVEN IF there is no lock file and the fuser command comes up empty. In that case, the only way to resolve the problem is by rebooting."
If you can't establish communication through a serial port to a device that is not asserting CD (Carrier Detect), try setting the environment variable "ttya-ignore-cd" to "true" (replace "ttya" with the port name).
Current advice from Sun is to always the /dev/cua/x devices for dialing out, rather than the /dev/term/x. Nevertheless, if you have trouble dialing out with one, try the other.
Reportedly, if you start C-Kermit and "set line" to a port that has a modem connected to it that is not turned on, and then "set flow rts/cts", there might be some (unspecified) difficulties closing the device because the CTS signal is not coming in from the modem.
The built-in SunLink X.25 support for Solaris 2.3/2.4./25 and SunLink 8.01 or 9.00 works OK provided the X.25 system has been installed and initialized properly. Packet sizes might need to be reduced to 256, maybe even less, depending on the configuration of the X.25 installation. On one connection where C-Kermit 6.0 was tested, very large packets and window sizes could be used in one direction, but only very small ones would work in the other.
In any case, according to Sun, C-Kermit's X.25 support is superfluous with SunLink 8.x / Solaris 2.3. Quoting an anonymous Sun engineer:
... there is now no need to include any X.25 code within kermit. As of X.25 8.0.1 we support the use of kermit, uucp and similar protocols over devices of type /dev/xty. This facility was there in 8.0, and should also work on the 8.0 release if patch 101524 is applied, but I'm not 100% sure it will work in all cases, which is why we only claim support from 8.0.1 onwards.When configuring X.25, on the "Advanced Configuration->Parameters" screen of the x25tool you can select a number of XTY devices. If you set this to be > 1, press Apply, and reboot, you will get a number of /dev/xty entries created.
Ignore /dev/xty0, it is a special case. All the others can be used exactly as if they were a serial line (e.g. /dev/tty) connected to a modem, except that instead of using Hayes-style commands, you use PAD commands.
From kermit you can do a 'set line' command to, say, /dev/xty1, then set your dialing command to be "CALL 12345678", etc. All the usual PAD commands will work (SET, PAR, etc).
I know of one customer in Australia who is successfully using this, with kermit scripts, to manage some X.25-connected switches. He used standard kermit, compiled for Solaris 2, with X.25 8.0 xty devices.
Hints for using a Sun workstation keyboard for VT emulation when accessing VMS, from the comp.os.vms newsgroup:
From: Jerry Leichter <leichter@smarts.com>
Newsgroups: comp.os.vms
Subject: Re: VT100 keyboard mapping to Sun X server
Date: Mon, 19 Aug 1996 12:44:21 -0400> I am stuck right now using a Sun keyboard (type 5) on systems running SunOS
> and Solaris. I would like to use EVE on an OpenVMS box with display back to
> the Sun. Does anyone know of a keyboard mapping (or some other procedure)
> which will allow the Sun keyboard to approximate a VT100/VT220?You can't get it exactly - because the keypad has one fewer key - but you can come pretty close. Here's a set of keydefs I use:
keycode 101=KP_0 keycode 119=KP_1 keycode 120=KP_2 keycode 121=KP_3 keycode 98=KP_4 keycode 99=KP_5 keycode 100=KP_6 keycode 75=KP_7 keycode 76=KP_8 keycode 77=KP_9 keycode 52=KP_F1 keycode 53=KP_F2 keycode 54=KP_F3 keycode 57=KP_Decimal keycode 28=Left keycode 29=Right keycode 30=KP_Separator keycode 105=KP_F4 keycode 78=KP_Subtract keycode 8=Left keycode 10=Right keycode 32=Up keycode 33=Down keycode 97=KP_EnterPut this in a file - I use "keydefs" in my home directory and feed it into xmodmap:
xmodmap - <$HOME/keydefsThis takes care of the arrow keys and the "calculator" key cluster. The "+" key will play the role of the DEC "," key. The Sun "-" key will be like the DEC "-" key, though it's in a physically different position - where the DEC PF4 key is. The PF4 key is ... damn, I'm not sure where "key 105" is. I *think* it may be on the leftmost key of the group of four just above the "calculator" key cluster.
I also execute the following (this is all in my xinitrc file):
xmodmap -e 'keysym KP_Decimal = KP_Decimal' xmodmap -e 'keysym BackSpace = Delete BackSpace' \ -e 'keysym Delete = BackSpace Delete' xmodmap -e 'keysym KP_Decimal = Delete Delete KP_Decimal' xmodmap -e 'add mod1 = Meta_R' xmodmap -e 'add mod1 = Meta_L'Beware of one thing about xmodmap: Keymap changes are applied to the *whole workstation*, not just to individual windows. There is, in fact, no way I know of to apply them to individual windows. These definitions *may* confuse some Unix programs (and/or some Unix users).
If you're using Motif, you may also need to apply bindings at the Motif level. If just using xmodmap doesn't work, I can try and dig that stuff up for you.
The following is a report from a user of C-Kermit 8.0 on Solaris 8 and 9, who had complained that while Kermit file transfers worked perfectly on direct (non-PPP) dialout connections, they failed miserably on PPP connections. We suggested that the PPP dialer probably was not setting the port and/or modem up in the same way that Kermit did:
I want to get back on this and tell you what the resolution was. You pointed me in the direction of flow control, which turned out to be the key.Some discussion on the comp.unix.solaris newsgroup led to some comments from Greg Andrews about the need to use the uucp driver to talk to the modem (/dev/cua/a). I had to remind Greg that no matter what the manpages for the zs and se drivers say, the ppp that Sun released with Solaris 8 7/01, and has in Solaris 9, is a setuid root program, and simply trying to make a pppd call from user space specifying /dev/cua/a would fail because of permissions. Greg finally put the question to the ppp people, who came back with information that is not laid out anywhere in the docs available for Solaris users. Namely, put /dev/cua/a in one of the priviledged options files in the /etc/ppp directory. That, plus resetting the OBP ttya-ignore-cd flag (this is Sun hardware) to false, seems to have solved the problems.
While I note that I had installed Kermit suid to uucp to use /dev/cua/a on this particular box, it seems to run fine through /dev/term/a. Not so with pppd.
With this change in place, I seem to be able to upload and download through telnet run on Kermit with the maximum length packets. I note that the window allocation display does show STREAMING, using telnet. Running ssh on Kermit, I see the standard 1 of 30 windows display, and note that there appears to be a buffer length limit between 1000 and 2000 bytes. Run with 1000, and it's tick-tock, solid as a rock. With 2000 I see timeout errors and RTS/CTS action on the modem.
Kermit's packet-length and other controls let you make adjustments like this to get around whatever obstacles might be thrown up -- in this case (running Kermit over ssh), the underling Solaris PTY driver.
C-Kermit can't be compiled successfully under Solaris 2.3 using SUNWspro cc 2.0.1 unless at least some of the following patches are applied to cc (it is not known which one(s), if any, fix the problem):
With unpatched cc 2.0.1, the symptom is that certain modules generate truncated object files, resulting in many unresolved references at link time.
The rest of the problems in this section have to do with bidirectional terminal ports and the Solaris Port Monitor. A bug in C-Kermit 5A ticked a bug in Solaris. The C-Kermit bug was fixed in version 6.0, and the Solaris bug was fixed in 2.4 (I think, or maybe 2.5).
Reportedly, "C-Kermit ... causes a SPARCstation running Solaris 2.3 to panic after the modem connects. I have tried compiling C-Kermit with Sun's unbundled C compiler, with GCC Versions 2.4.5 and 2.5.3, with make targets 'sunos51', 'sunos51tcp', 'sunos51gcc', and even 'sys5r4', and each time it compiles and starts up cleanly, but without fail, as soon as I dial the number and get a 'CONNECT' message from the modem, I get:
BAD TRAP kermit: Data fault kernel read fault at addr=0x45c, pme=0x0 Sync Error Reg 80 <INVALID> ... panic: Data Fault. ... Rebooting...
The same modem works fine for UUCP/tip calling." Also (reportedly), this only happens if the dialout port is configured as in/out via admintool. If it is configured as out-only, no problem. This is the same dialing code that works on hundreds of other System-V based Unix OS's. Since it should be impossible for a user program to crash the operating system, this problem must be chalked up to a Solaris bug. Even if you SET CARRIER OFF, CONNECT, and dial manually by typing ATDTnnnnnnn, the system panics as soon as the modem issues its CONNECT message. (Clearly, when you are dialing manually, C-Kermit does not know a thing about the CONNECT message, and so the panic is almost certainly caused by the transition of the Carrier Detect (CD) line from off to on.) This problem was reported by many users, all of whom say that C-Kermit worked fine on Solaris 2.1 and 2.2. If the speculation about CD is true, then a possible workaround might be to configure the modem to leave CD on (or off) all the time. Perhaps by the time you read this, a patch will have been issued for Solaris 2.3.
The following is from Karl S. Marsh, Systems & Networks Administrator, AMBIX Systems Corp, Rochester, NY:
Environment: Solaris 2.3 Patch 101318-45 C-Kermit 5A(189) (and presumably this applies to 188 and 190 also). eeprom setting:
ttya-rts-dtr-off=false ttya-ignore-cd=false ttya-mode=19200,8,n,8,-To use C-Kermit on a bidirectional port in this environment, do not use admintool to configure the port. Use admintool to delete any services running on the port and then quit admintool and issue the following command:
pmadm -a -p zsmon -s ttyb -i root -fu -v 1 -m "`ttyadm -b -d /dev/term/b \ -l conttyH -m ldterm,ttcompat -s /usr/bin/login -S n`"[NOTE: This was copied from a blurry fax, so please check it carefully] where:
-a = Add service -p = pmtag (zsmon) -s = service tag (ttyb) -i = id to be associated with service tag (root) -fu = create utmp entry -v = version of ttyadm -m = port monitor-specific portion of the port monitor administrative file entry for the service -b = set up port for bidirectional use -d = full path name of device -l = which ttylabel in the /etc/ttydefs file to use -m = a list of pushable STREAMS modules -s = pathname of service to be invoked when connection request received -S = software carrier detect on or off (n = off)"This is exactly how I was able to get Kermit to work on a bi-directional port without crashing the system."
On the Solaris problem, also see SunSolve Bug ID 1150457 ("Using C-Kermit, get Bad Trap on receiving prompt from remote system"). Another user reported "So, I have communicated with the Sun tech support person that submitted this bug report [1150457]. Apparently, this bug was fixed under one of the jumbo kernel patches. It would seem that the fix did not live on into 101318-45, as this is EXACTLY the error that I see when I attempt to use kermit on my system."
Later (Aug 94)... C-Kermit dialout successfully tested on a Sun4m with a heavily patched Solaris 2.3. The patches most likely to have been relevant:
Still later (Nov 94): another user (Bo Kullmar in Sweden) reports that after using C-Kermit to dial out on a bidirectional port, the port might not answer subsequent incoming calls, and says "the problem is easy enough to fix with the Serial Port Manager; I just delete the service and install it again using the graphical interface, which underneath uses commands like sacadm and pmadm." Later Bo reports, "I have found that if I run Kermit with the following script then it works. This script is for /dev/cua/a, "-s a" is the last a in /dev/cua/a:
#! /bin/sh kermit sleep 2 surun pmadm -e -p zsmon -s a
For additional information, see "Celeste's Tutorial on SunOS 4.1.3+ Modems and Terminals":
http://www.stokely.com/
For FAQs, etc, from Sun, see:
For history of Sun models and SunOS versions, see (should be all the same):
Sun SPARCstation users should read the section "Setting up Modem Software" in the Desktop SPARC Sun System & Network Manager's Guide. If you don't set up your serial ports correctly, Kermit (and other communications software) won't work right.
Also, on certain Sun models like IPC, the serial port hardware might need to have a jumper changed to make it an RS-232 port rather than RS-423.
Reportedly, C-Kermit does not work correctly on a Sun SPARCstation in an Open Windows window with scrolling enabled. Disable scrolling, or else invoke Kermit in a terminal emulation window (xterm, crttool, vttool) under SunView (this might be fixed in later SunOS releases).
On the Sun with Open Windows, an additional symptom has been reported: outbound SunLink X.25 connections "magically" translate CR typed at the keyboard into LF before transmission to the remote host. This doesn't happen under SunView.
SET CARRIER ON, when used on the SunOS 4.1 version of C-Kermit (compiled in the BSD universe), causes the program to hang uninterruptibly when SET LINE is issued for a device that is not asserting carrier. When Kermit is built in the Sys V universe on the same computer, there is no problem (it can be interrupted with Ctrl-C). This is apparently a limitation of the BSD-style tty driver.
SunOS 4.1 C-Kermit has been observed to dump core when running a complicated script program under cron. The dump invariably occurs in ttoc(), while trying to output a character to a TCP/IP TELNET connection. ttoc() contains a write() call, and when the system or the network is very busy, the write() call can get stuck for long periods of time. To break out of deadlocks caused by stuck write() calls, there is an alarm around the write(). It is possible that the core dump occurs when this alarm signal is caught. (This one has not been observed recently -- possibly fixed in edit 190.)
On Sun computers with SunOS 4.0 or 4.1, SET FLOW RTS/CTS works only if the carrier signal is present from the communication device at the time when C-Kermit enters packet mode or CONNECT mode. If carrier is not sensed (e.g. when dialing), C-Kermit does not attempt to turn on RTS/CTS flow control. This is because the SunOS serial device driver does not allow characters to be output if RTS/CTS is set (CRTSCTS) but carrier (and DSR) are not present. Workaround (maybe): SET CARRIER OFF before giving the SET LINE command, establish the connection, then SET FLOW RTS/CTS
It has also been reported that RTS/CTS flow control under SunOS 4.1 through 4.1.3 works only on INPUT, not on output, and that there is a patch from Sun to correct this problem: Patch-ID# T100513-04, 20 July 1993 (this patch might apply only to SunOS 4.1.3). It might also be necessary to configure the eeprom parameters of the serial port; e.g. do the following as root at the shell prompt:
eeprom ttya-ignore-cd=false eeprom ttya-rts-dtr-off=true
There have been reports of file transfer failures on Sun-3 systems when using long packets and/or large window sizes. One user says that when this happens, the console issues many copies of this message:
chaos vmunix: zs1: ring buffer overflow
This means that SunOS is not scheduling Kermit frequently enough to service interrupts from the zs serial device (Zilog 8350 SCC serial communication port) before its input silo overflows. Workaround: use smaller packets and/or a smaller window size, or use "nice" to increase Kermit's priority. Use hardware flow control if available, or remove other active processes before running Kermit.
SunLink X.25 support in C-Kermit 5A(190) was built and tested successfully under SunOS 4.1.3b and SunLink X.25 7.00.
See also: The comp.unix.ultrix and comp.sys.dec newsgroups.
There is no hardware flow control in Ultrix. That's not a Kermit deficiency, but an Ultrix one.
When sending files to C-Kermit on a Telnet connection to a remote Ultrix system, you must SET PREFIXING ALL (or at least prefix more control characters than are selected by SET PREFIXING CAUTIOUS).
Reportedly, DEC ULTRIX 4.3 is immune to C-Kermit's disabling of SIGQUIT, which is the signal that is generated when the user types Ctrl-\, which kills the current process (i.e. C-Kermit) and dumps core. Diagnosis and cure unknown. Workaround: before starting C-Kermit -- or for that matter, when you first log in because this applies to all processes, not just Kermit -- give the following Unix command:
stty quit undef
Certain operations driven by RS-232 modem signal do not work on DECstations or other DEC platforms whose serial interfaces use MMP connectors (DEC version of RJ45 telephone jack with offset tab). These connectors convey only the DSR and DTR modem signals, but not carrier (CD), RTS, CTS, or RI. Use SET CARRIER OFF to enable communication, or "hotwire" DSR to CD.
The maximum serial speed on the DECstation 5000 is normally 19200, but various tricks are available (outside Kermit) to enable higher rates. For example, on the 5000/200, 19200 can be remapped (somehow, something to do with "a bit in the SIR", whatever that is) to 38400, but in software you must still refer to this speed as 19200; you can't have 19200 and 38400 available at the same time.
19200, reportedly, is also the highest speed supported by Ultrix, but NetBSD reportedly supports speeds up to 57600 on the DECstation, although whether and how well this works is another question.
In any case, given the lack of hardware flow control in Ultrix, high serial speeds are problematic at best.
See also:
Also see general comments on PC-based Unixes in Section 3.0. By the way, this section is separate from the SCO (Caldera) section because at the time this section was started, Unixware was owned by a company called Univel. Later it was sold to Novell, and then to SCO. Still later, SCO was sold to Caldera.
In Unixware 2.0 and later, the preferred serial device names (drivers) are /dev/term/00 (etc), rather than /dev/tty00 (etc). Note the following correspondence of device names and driver characteristics:
New name Old name Description /dev/term/00 /dev/tty00 ??? /dev/term/00h /dev/tty00h Modem signals and hardware flow control /dev/term/00m /dev/tty00m Modem signals(?) /dev/term/00s /dev/tty00s Modem signals and software flow control /dev/term/00t /dev/tty00t ???
Lockfile names use device.major.minor numbers, e.g.:
/var/spool/locks/LK.7679.003.005
The minor number varies according to the device name suffix (none, h, m, s, or t). Only the device and major number are compared, and thus all of the different names for the same physical device (e.g. all of those shown in the table above) interlock effectively.
Prior to UnixWare 7, serial speeds higher than 38400 are not supported. In UnixWare 7, we also support 57600 and 115200, plus some unexpected ones like 14400, 28800, and 76800, by virtue of a strange new interface, evidently peculiar to UnixWare 7, discovered while digging through the header files: tcsetspeed(). Access to this interface is allowed only in POSIX builds, and thus the UnixWare 7 version of C-Kermit is POSIX-based, unlike C-Kermit for Unixware 1.x and 2.x (since the earlier UnixWare versions did not support high serial speeds, period).
HOWEVER, turning on POSIX features engages all of the "#if (!_POSIX_SOURCE)" clauses in the UnixWare header files, which in turn prevent us from having modem signals, access to the hardware flow control APIs, select(), etc -- in short, all the other things we need in communications software, especially when high speeds are used. Oh the irony. And so C-Kermit must be shamelessly butchered -- as it has been so many times before -- to allow us to have the needed features from the POSIX and non-POSIX worlds. See the UNIXWAREPOSIX sections of ckutio.c.
After the butchery, we wind up with Unixware 2.x having full modem-signal capability, but politically-correct Unixware 7.x lacking the ability to automatically detect a broken connection when carrier drops.
Meanwhile the Unixware tcsetspeed() function allows any number at all (any long, 0 or positive) as an argument and succeeds if the number is a legal bit rate for the serial device, and fails otherwise. There is no list anywhere of legal speeds. Thus the SET SPEED keyword table ("set speed ?" to see it) is hardwired based on trial and error with all known serial speeds, the maximum being 115200. However, to allow for the possibility that other speeds might be allowed in the future (or with different port drivers), the SET SPEED command for UnixWare 7 only allows you to specify any number at all; a warning is printed if the number is not in the list, but the number is accepted anyway; the command succeeds if tcsetspeed() accepts the number, and fails otherwise.
In C-Kermit 8.0 testing, it was noticed that the POSIX method for hanging up the phone by dropping DTR (set speed 0, pause, restore speed) did not actually drop DTR. The APIs do not return any error indication, but nothing happens. I changed tthang() to skip the special case I had made for Unixware and instead follow the normal path: if TIOCSDTR is defined use that, otherwise blah blah... It turns out TIOCSDTR *is* defined, and it works.
So in Unixware (at least in 2.1.3) we can read modem signals, hangup by toggling DTR, and so on, BUT... But once the remote hangs up and Carrier drops, the API for reading modem signals ceases to function; although the device is still open, the TIOCMGET ioctl always raises errno 6 = ENXIO, "No such device or address".
Old business:
Using C-Kermit 6.0 on the UnixWare 1.1 Application Server, one user reported a system panic when the following script program is executed:
set line /dev/tty4 set speed 9600 output \13 connect
The panic does not happen if a PAUSE is inserted:
set line /dev/tty4 set speed 9600 pause 1 output \13 connect
This is using a Stallion EasyIO card installed as board 0 on IRQ 12 on a Gateway 386 with the Stallion-supplied driver. The problem was reported to Novell and Stallion and (reportedly) is now fixed.
Reportedly, version 5A(190), when built under Apollo SR10 using "make sr10-bsd", compiles, links, and executes OK, but leaves the terminal unusable after it exits -- the "cs7" or "cs8" (character size) parameter has become cs5. The terminal must be reset from another terminal. Cause and cure unknown. Suggested workaround: Wrap Kermit in a shell script something like:
kermit @* stty sane
C-Kermit 7.0 was too big to be built on Tandy Xenix, even in a minimum configuration; version 6.0 is the last one that fits.
Reportedly, in C-Kermit 6.0, if you type lots of Ctrl-C's during execution of the initialization file, ghost Kermit processes will be created, and will compete for the keyboard. They can only be removed via "kill -9" from another terminal, or by rebooting. Diagnosis -- something strange happening with the SIGINT handler while the process is reading the directory (it seems to occur during the SET PROMPT [\v(dir)] ... sequence). Cure: unknown. Workaround: don't interrupt C-Kermit while it is executing its init file on the Tandy 16/6000.
While putting together and testing C-Kermit 8.0, it was discovered that binaries built for one version of Tru64 Unix (e.g. 4.0G) might exhibit very strange behavior if run on a different version of Tru64 Unix (e.g. 5.1A). The typical symptom was that a section of the initialization file would be skipped, notably locating the dialing and/or network directory as well as finding and executing the customization file, ~/.mykermrc. This problem also is reported to occur on Tru64 Unix 5.0 (Rev 732) even when running a C-Kermit binary that was built there. However, the Tru64 5.1A binary works correctly on 5.0. Go figure.
When making Telnet connections to a Digital Unix or Tru64 system, and your Telnet client forwards your user name, the Telnet server evidently stuffs the username into login's standard input, and you see:
login: ivan Password:
This is clearly going to play havoc with scripts that look for "login:". Workaround (when Kermit is your Telnet client): SET LOGIN USER to nothing, to prevent Kermit from sending your user ID.
Before you can use a serial port on a new Digital Unix system, you must run uucpsetup to enable or configure the port. Evidently the /dev/tty00 and 01 devices that appear in the configuration are not usable; uucpsetup turns them into /dev/ttyd00 and 01, which are. Note that uucpsetup and other uucp-family programs are quite primitive -- they only know about speeds up to 9600 bps and their selection of modems dates from the early 1980s. None of this affects Kermit, though -- with C-Kermit, you can use speeds up to 115200 bps (at least in DU4.0 and later) and modern modems with hardware flow control and all the rest.
Reportedly, if a modem is set for &S0 (assert DSR at all times), the system resets or drops DTR every 30 seconds; reportedly DEC says to set &S1.
Digital Unix 3.2 evidently wants to believe your terminal is one line longer than you say it is, e.g. when a "more" or "man" command is given. This is has nothing to do with C-Kermit, but tends to annoy those who use Kermit or other terminal emulators to access Digital Unix systems. Workaround: tell Unix to "stty rows 23" (or whatever).
Reportedly, there is some bizarre behavior when trying to use a version of C-Kermit built on one Digital Unix 4.0 system on another one, possibly due to differing OS or library revision levels; for example, the inability to connect to certain TCP/IP hosts. Solution: rebuild C-Kermit from source code on the system where you will be using it.
Digital Unix tgetstr() causes a segmentation fault. C-Kermit 7.0 added #ifdefs to avoid calling this routine in Digital Unix. As a result, the SCREEN commands always send ANSI escape sequences -- even though curses knows your actual terminal type.
Reportedy the Tru64 Unix 4.0E 1091 Telnet server does not tolerate streaming transfers into itself, at least not when the sending Kermit is on the same local network. Solution: tell one Kermit or the other (or both) to "set streaming off". This might or might be the case with earlier and/or later Tru64, Digital Unix, and OSF/1 releases.
See also:
About IRIX version numbers: "uname -a" tells the "two-digit" version number, such as "5.3" or "6.5". The three-digit form can be seen with "uname -R". (this information is unavailable at the simple API level). Supposedly all three-digit versions within the same two-digit version (e.g. 6.5.2, 6.5.3) are binary compatible; i.e. a binary built on any one of them should run on all others. The "m" suffix denotes just patches; the "f" suffix indicates that features were added.
An IRIX binary built on lower MIPS model (Instruction Set Architecture, ISA) can run on higher models, but not vice versa:
Furthermore, there are different Application Binary Inferfaces (ABIs):
MIPS1 R3000 and below MIPS2 R4000 MIPS3 R4x00 MIPS4 R5000 and above
COFF 32 bits, IRIX 5.3, 5.2, 5.1, 4.x and below o32 ELF 32 bits, IRIX 5.3, 6.0 - 6.5 N32 ELF 32 bits, IRIX 6.2 - 6.5 N64 ELF 64 bits, IRIX 6.2 - 6.5
Thus a prebuilt IRIX binary works on a particular machine only if (a) the machine's IRIX version (to one decimal place) is equal to or greater than the version under which the binary was built; (b) the machine's MIPS level is greater or equal to that of the binary; and (c) the machine supports the ABI of the binary. If all three conditions are not satisfied, of course, you can build a binary yourself from source code since, unlike some other Unix vendors, SGI does supply a C compiler and libraries.
SGI did not supply an API for hardware flow control prior to IRIX 5.2. C-Kermit 6.1 and higher for IRIX 5.2 and higher supports hardware flow control in the normal way, via "set flow rts/cts".
For hardware flow control on earlier IRIX and/or C-Kermit versions, use the ttyf* (modem control AND hardware flow control) devices and not the ttyd* (direct) or ttym* (modem control but no hardware flow control) ones, and obtain the proper "hardware handshaking" cable from SGI, which is incompatible with the ones for the Macintosh and NeXT even though they look the same ("man serial" for further info) and tell Kermit to "set flow keep" and "set modem flow rts/cts".
Serial speeds higher than 38400 are available in IRIX 6.2 and later, on O-class machines (e.g. Origin, Octane) only, and are supported by C-Kermit 7.0 and later. Commands such as "set speed 115200" may be given on other models (e.g. Iris, Indy, Indigo) but will fail because the OS reports an invalid speed for the device.
Experimentation with both IRIX 5.3 and 6.2 shows that when logged in to IRIX via Telnet, that remote-mode C-Kermit can't send files if the packet length is greater than 4096; the Telnet server evidently has this restriction (or bug), since there is no problem sending long packets on serial or rlogin connections. However, it can receive files with no problem if the packet length is greater than 4096. As a workaround, the FAST macro for IRIX includes "set send packet-length 4000". IRIX 6.5.1 does not have this problem, so evidently it was fixed some time after IRIX 6.2. Tests show file-transfer speeds are better (not worse) with 8K packets than with 4K packets from IRIX 6.5.1.
Reportedly some Indys have bad serial port hardware. IRIX 5.2, for example, needs patch 151 to work around this; or upgrade to a later release. Similarly, IRIX 5.2 has several problems with serial i/o, flow control, etc. Again, patch or upgrade.
Reportedly on machines with IRIX 4.0, Kermit cannot be suspended by typing the suspend ("swtch") character if it was started from csh, even though other programs can be suspended this way, and even though the Z and SUSPEND commands still work correctly. This is evidently because IRIX's csh does not deliver the SIGTSTP signal to Kermit. The reason other programs can be suspended in the same environment is probably that they do not trap SIGTSTP themselves, so the shell is doing the suspending rather than the application.
Also see notes about IRIX 3.x in the C-Kermit for Unix Installation Instructions.
If you have problems making TCP/IP connections in versions of IRIX built with GCC 2.95.2, see the bugs section of:
http://freeware.sgi.com/Installable/gcc-2.95.2.html.
Reportedly, if you allow gcc to compile C-Kermit on Irix you should be aware that there might be problems with some of the network code. The specifics are at http://freeware.sgi.com/Installable/gcc-2.95.2.html; scroll down to the "known bugs" section at the end of the document.
See also: The comp.sys.be newsgroup.
The BeBox has been discontinued and BeOS repositioned for PC platforms. The POSIX parts of BeOS are not finished, nor is the sockets library, therefore a fully functional version of C-Kermit is not possible. In version 6.0 of C-Kermit, written for BeOS DR7, it was possible to:
The following do not work:
C-Kermit does not work on BeOS DR8 because of changes in the underlying APIs. Unfortunately not enough changes were made to allow the regular POSIX-based C-Kermit to work either. Note: the lack of a fork() service requires the select()-based CONNECT module, but there is no select(). There is a select() in DR8, but it doesn't work.
C-Kermit 7.0 was built for BeOS 4.5 and works in remote mode. It does not include networking support since the APIs are still not there. It is not known if dialing out works, but probably not. Be experts are welcome to lend a hand.
Somebody downloaded the C-Kermit 6.0 binary built under DG/UX 5.40 and ran it under DG/UX 5.4R3.10 -- it worked OK except that file dates for incoming files were all written as 1 Jan 1970. Cause and cure unknown. Workaround: SET ATTRIBUTE DATE OFF. Better: Use a version of C-Kermit built under and for DG/UX 5.4R3.10.
Reportedly, when coming into a Sequent Unix (DYNIX) system through an X.25 connection, Kermit doesn't work right because the Sequent's FIONREAD ioctl returns incorrect data. To work around, use the 1-character-at-a-time version of myread() in ckutio.c (i.e. undefine MYREAD in ckutio.c and rebuild the program). This is unsatisfying because two versions of the program would be needed -- one for use over X.25, and the other for serial and TCP/IP connections.
Some NebBSD users have reported difficulty escaping back from CONNECT mode, usually when running NetBSD on non-PC hardware. Probably a keyboard issue.
NetBSD users have also reported that C-Kermit doesn't pop back to the prompt if the modem drops carrier. This needs to be checked out & fixed if possible.
(All the above seems to work properly in C-Kermit 7.0 and later.)
Nomenclature:
mac os x server 1.0 = rhapsody 5.3 mac os x server 1.0.1 = rhapsody 5.4 mac os x server 1.0.2 = rhapsody 5.5 mac os x server 1.2 = rhapsody 5.6 mac os x 10.0 and 10.0.x run on darwin 1.3.y mac os x 10.1 and 10.1.x run on darwin 1.4.y mac os x (client) to darwin =(approximately)= X11 to unix
Evidently hardware flow control doesn't work. This has been reported to Apple (June 2001), no response yet that I know of.
Also see:
Mark Williams COHERENT was perhaps the first commercial Unix-based operating system for PCs, first appearing about 1983 or -84 for the PC/XT (?), and popular until about 1993, when Linux took over. C-Kermit, as of version 8.0, is still current for COHERENT 386 4.2 (i.e. only for i386 and above). Curses is included, but lots of other features are omitted due to lack of the appropriate OS features, APIs, libraries, hardware, or just space: e.g. TCP/IP, floating-point arithmetic, learned scripts. Earlier versions of COHERENT ran on 8086 and 80286, but these are to small to build or run C-Kermit, but G-Kermit should be OK (as might be ancient versions of C-Kermit).
You can actually build a version with floating point support -- just take -DNOFLOAT out of CFLAGS and add -lm to LIBS; NOFLOAT is the default because COHERENT tends to run on old PCs that don't have floating-point hardware. You can also add "-f" to CFLAGS to have it link in the floating-point emulation library. Also I'm not sure why -DNOLEARN is included, since it depends on select(), which COHERENT has.
[/usr/olga] C-Kermit>
(In C-Kermit 7.0 the square braces were replaced by round parentheses to avoid conflicts with ISO 646 national character sets.)
If that directory is on an NFS-mounted disk, and NFS stops working or the disk becomes unavailable, C-Kermit will hang waiting for NFS and/or the disk to come back. Whether you can interrupt C-Kermit when it is hung this way depends on the specific OS. Kermit has called the operating systems's getcwd() function, and is waiting for it to return. Some versions of Unix (e.g. HP-UX 9.x) allow this function to be interrupted with SIGINT (Ctrl-C), others (such as HP-UX 8.x) do not. To avoid this effect, you can always use SET PROMPT to change your prompt to something that does not involve calling getcwd(), but if NFS is not responding, C-Kermit will still hang any time you give a command that refers to an NFS-mounted directory. Also note that in some cases, the uninterruptibility of NFS-dependent system or library calls is considered a bug, and sometimes there are patches. For HP-UX, for example:
replaced by:
HP-UX 10.20 libc PHCO_8764 PHCO_14891/PHCO_16723
HP-UX 10.10 libc PHCO_8763 PHCO_14254/PHCO_16722
HP-UX 9.x libc PHCO_7747 S700 PHCO_13095
HP-UX 9.x libc PHCO_6779 S800 PHCO_11162
The same can apply to any other environment in which the user's session is captured, monitored, recorded, or manipulated. Examples include the 'script' program (for making a typescript of a session), the Computronics PEEK package and pksh (at least versions of it prior to 1.9K), and so on.
You might try the following -- what we call "doomsday Kermit" -- settings to push packets through even the densest and most obstructive connections, such as "screen" and "splitvt" (and certain kinds of 3270 protocol emulators): Give these commands to BOTH Kermit programs:
SET FLOW NONE SET CONTROL PREFIX ALL SET RECEIVE PACKET-LENGTH 70 SET RECEIVE START 62 SET SEND START 62 SET SEND PAUSE 100 SET BLOCK B
If it works, it will be slow.
C-Kermit> DATE 20011028 05:01:02 GMT ; EDT 20011028 01:01:02 C-Kermit> DATE 20011028 06:01:02 GMT ; EST 20011028 01:01:02 C-Kermit>
but the implicit change in timezone offset is not recognized:
C-Kermit> echo \fdiffdate(20011028 05:01:02 GMT, 20011028 06:01:02 GMT) +0:00 C-Kermit>Date/time arithmetic, offsets, delta times, and timezone support are new to C-Kermit 8.0, and might be expected to evolve and improve in subsequent releases.
On some platforms, files downloaded with HTTP receive the current timestamp, rather than the HTTP "Last Modified" time (this can be fixed by including utime.h, e.g. in SunOS and Tru64...).
SSH and PTY commands can fail if (a) all pseudoterminals are in use; or (b) you do not have read/write access to the pseudoterminal that was assigned. An example of (b) was reported with the Zipslack Slackware Linux distribution, in which the pseudoterminals were created with crw-r--r-- permission, instead of crw-rw-rw-.
C-Kermit's initialization file for Unix is .kermrc (lowercase, starts with period) in your home directory, unless Kermit was built with the system-wide initialization-file option (see the C-Kermit for Unix Installation Instructions).
C-Kermit identifies your home directory based on the environment variable, HOME. Most Unix systems set this variable automatically when you log in. If C-Kermit can't find your initialization file, check your HOME variable:
echo $HOME (at the Unix prompt)
or:
echo \$(HOME) (at the C-Kermit prompt)
If HOME is not defined, or is defined incorrectly, add the appropriate definition to your Unix .profile or .login file, depending on your shell:
setenv HOME full-pathname-of-your-home-directory (C-Shell, .login file)
or:
HOME=full-pathname-of-your-home-directory (sh, ksh, .profile file) export HOME
NOTE: Various other operations depend on the correct definition of HOME. These include the "tilde-expansion" feature, which allows you to refer to your home directory as "~" in filenames used in C-Kermit commands, e.g.:
send ~/.kermrc
as well as the \v(home) variable.
Prior to version 5A(190), C-Kermit would look for its initialization file in the current directory if it was not found in the home directory. This feature was removed from 5A(190) because it was a security risk. Some people, however, liked this behavior and had .kermrc files in all their directories that would set up things appropriately for the files therein. If you want this behavior, you can accomplish it in various ways, for example:
alias kd="kermit -Y ./.kermrc"
take ./.kermrc
Suppose you need to pass a password from the Unix command line to a C-Kermit script program, in such a way that it does not show up in "ps" or "w" listings. Here is a method (not guaranteed to be 100% secure, but definitely more secure than the more obvious methods):
echo mypassword | kermit myscript
The "myscript" file contains all the commands that need to be executed during the Kermit session, up to and including EXIT, and also includes an ASK or ASKQ command to read the password from standard input, which has been piped in from the Unix 'echo' command, but it must not include a CONNECT command. Only "kermit myscript" shows up in the ps listing.
Version-7 based Unix implementations, including 4.3 BSD and earlier and Unix systems based upon BSD, use a 4-bit field to record a serial device's terminal speed. This leaves room for 16 speeds, of which the first 14 are normally:
0, 50, 75, 110, 134.5, 150, 200, 300, 600, 1200, 1800, 2400, 4800, and 9600
The remaining two are usually called EXTA and EXTB, and are defined by the particular Unix implementation. C-Kermit determines which speeds are available on your system based on whether symbols for them are defined in your terminal device header files. EXTA is generally assumed to be 19200 and EXTB 38400, but these assumptions might be wrong, or they might not apply to a particular device that does not support these speeds. Presumably, if you try to set a speed that is not legal on a particular device, the driver will return an error, but this can not be guaranteed.
On these systems, it is usually not possible to select a speed of 14400 bps for use with V.32bis modems. In that case, use 19200 or 38400 bps, configure your modem to lock its interface speed and to use RTS/CTS flow control, and tell C-Kermit to SET FLOW RTS/CTS and SET DIAL SPEED-MATCHING OFF.
The situation is similar, but different, in System V. SVID Third Edition lists the same speeds, 0 through 38400.
Some versions of Unix, and/or terminal device drivers that come with certain third-party add-in high-speed serial communication interfaces, use the low "baud rates" to stand for higher ones. For example, SET SPEED 50 gets you 57600 bps; SET SPEED 75 gets you 76800; SET SPEED 110 gets 115200.
SCO ODT 3.0 is an example where a "baud-rate-table patch" can be applied that can rotate the tty driver baud rate table such that 600=57600 and 1800=115k baud. Similarly for Digiboard multiport/portservers, which have a "fastbaud" setting that does this. Linux has a "setserial" command that can do it, etc.
More modern Unixes support POSIX-based speed setting, in which the selection of speeds is not limited by a 4-bit field. C-Kermit 6.1 incorporates a new mechanism for finding out (at compile time) which serial speeds are supported by the operating system that does not involve editing of source code by hand; on systems like Solaris 5.1, IRIX 6.2, and SCO OSR5.0.4, "set speed ?" will list speeds up to 460800 or 921600. In C-Kermit 7.0 and later:
When Kermit is given a "set speed" command for a particular device, the underlying system service is called to set the speed; its return code is checked and the SET SPEED command fails if the return code indicates failure. Regardless of the system service return status, the device's speed is then read back and if it does not match the speed that was requested, an error message is printed and the command fails.
Even when the command succeeds, this does not guarantee successful operation at a particular speed, especially a high one. That depends on electricity, information theory, etc. How long is the cable, what is its capacitance, how well is it shielded, etc, not to mention that every connection has two ends and its success depends on both of them. (With the obvious caveats about internal modems, is the cable really connected, interrupt conflicts, etc etc etc).
Note, in particular, that there is a certain threshold above which modems can not "autobaud" -- i.e. detect the serial interface speed when you type AT (or whatever else the modem's recognition sequence might be). Such modems need to be engaged at a lower speed (say 2400 or 9600 or even 115200 -- any speed below their autobaud threshold) and then must be given a modem-specific command (which can be found in the modem manual) to change their interface speed to the desired higher speed, and then the software must also be told to change to the new, higher speed.
For additional information, read Section 9.5 of the Installation Instructions, plus any platform-specific notes in Section 3 above.
Similarly, if you give a SET MODEM TYPE HAYES (or USR, or any other modem type besides DIRECT, NONE, or UNKNOWN) and then SET LINE to an empty port, the subsequent close (implicit or explicit) is liable to hang or even crash (through no fault of Kermit's -- the hanging or crashing is inside a system call such as cfsetospeed() or close()).
The SET CARRIER-WATCH command works as advertised only if the underlying operating system and device drivers support this feature; in particular only if a read() operation returns immediately with an error code if the carrier signal goes away or, failing that, if C-Kermit can obtain the modem signals from the device driver (you can tell by giving a "set line" command to a serial device, and then a "show communications" command -- if modem signals are not listed, C-Kermit won't be able to detect carrier loss, the WAIT command will not work, etc). Of course, the device itself (e.g. modem) must be configured appropriately and the cables convey the carrier and other needed signals, etc.
If you dial out from Unix system, but then notice a lot of weird character strings being stuck into your session at random times (especially if they look like +++ATQ0H0 or login banners or prompts), that means that getty is also trying to control the same device. You'll need to dial out on a device that is not waiting for a login, or else disable getty on the device.
As of version 7.0, C-Kermit makes explicit checks for the Carrier Detect signal, and so catches hung-up connections much better than 6.0 and earlier. However, it still can not be guaranteed to catch every ever CD on-to-off transition. For example, when the HP-UX version of C-Kermit is in CONNECT mode on a dialed connection and CARRIER-WATCH ON or AUTO, and you turn off the modem, HP-UX is stuck in a read() that never returns. (C-Kermit does not pop back to its prompt automatically, but you can still escape back.)
If, on the other hand, you log out from the remote system, and it hangs up, and CD drops on the local modem, C-Kermit detects this and pops back to the prompt as it should. (Evidently there can be a difference between CD and DSR turning off at the same time, versus CD turning off while DSR stays on; experimentation with &S0/&S1/&S2 on your modem might produce the desired results).
When Unix C-Kermit exits, it closes (and must close) the communications device. If you were dialed out, this will most likely hang up the connection. If you want to get out of Kermit and still use Kermit's communication device, you have several choices:
If you are having trouble dialing:
Make sure your dialout line is correctly configured for dialing out (as opposed to login). The method for doing this is different for each kind of Unix system. Consult your system documentation for configuring lines for dialing out (for example, Sun SparcStation IPC users should read the section "Setting up Modem Software" in the Desktop SPARC Sun System & Network Manager's Guide; HP-9000 workstation users should consult the manual Configuring HP-UX for Peripherals, etc).
Symptom: DIAL works, but a subsequent CONNECT command does not. Diagnosis: the modem is not asserting Carrier Detect (CD) after the connection is made, or the cable does not convey the CD signal. Cure: Reconfigure the modem, replace the cable. Workaround: SET CARRIER OFF (at least in System-V based Unix versions).
For Berkeley-Unix-based systems (4.3BSD and earlier), Kermit includes code to use LPASS8 mode when parity is none, which is supposed to allow 8-bit data and Xon/Xoff flow control at the same time. However, as of edit 174, this code is entirely disabled because it is unreliable: even though the host operating system might (or might not) support LPASS8 mode correctly, the host access protocols (terminal servers, telnet, rlogin, etc) generally have no way of finding out about it and therefore render it ineffective, causing file transfer failures. So as of edit 174, Kermit once again uses rawmode for 8-bit data, and so there is no Xon/Xoff flow control during file transfer or terminal emulation in the Berkeley-based versions (4.3 and earlier, not 4.4).
Also on Berkeley-based systems (4.3 and earlier), there is apparently no way to configure a dialout line for proper carrier handling, i.e. ignore carrier during dialing, require carrier thereafter, get a fatal error on any attempt to read from the device after carrier drops (this is handled nicely in System V by manipulation of the CLOCAL flag). The symptom is that carrier loss does not make C-Kermit pop back to the prompt automatically. This is evident on the NeXT, for example, but not on SunOS, which supports the CLOCAL flag. This is not a Kermit problem, but a limitation of the underlying operating system. For example, the cu program on the NeXT doesn't notice carrier loss either, whereas cu on the Sun does.
On certain AT&T Unix systems equipped with AT&T modems, DIAL and HANGUP don't work right. Workarounds: (1) SET DIAL HANGUP OFF before attempting to dial; (2) If HANGUP doesn't work, SET LINE, and then SET LINE <device> to totally close and reopen the device. If all else fails, SET CARRIER OFF.
C-Kermit does not contain any particular support for AT&T DataKit devices. You can use Kermit software to dial in to a DataKit line, but C-Kermit does not contain the specialized code required to dial out from a DataKit line. If the Unix system is connected to DataKit via serial ports, dialout should work normally (e.g. set line /dev/ttym1, set speed 19200, connect, and then see the DESTINATION: prompt, from which you can connect to another computer on the DataKit network or to an outgoing modem pool, etc). But if the Unix system is connected to the DataKit network through the special DataKit interface board, then SET LINE to a DataKit pseudodevice (such as /dev/dk031t) will not work (you must use the DataKit "dk" or "dkcu" program instead). In C-Kermit 7.0 and later, you can make Kermit connections "though" dk or dkcu using "set line /pty".
In some BSD-based Unix C-Kermit versions, SET LINE to a port that has nothing plugged in to it with SET CARRIER ON will hang the program (as it should), but it can't be interrupted with Ctrl-C. The interrupt trap is correctly armed, but apparently the Unix open() call cannot be interrupted in this case. When SET CARRIER is OFF or AUTO, the SET LINE will eventually return, but then the program hangs (uninterruptibly) when the EXIT or QUIT command (or, presumably, another SET LINE command) is given. The latter is probably because of the attempt to hang up the modem. (In edit 169, a timeout alarm was placed around this operation.)
With SET DIAL HANGUP OFF in effect, the DIAL command might work only once, but not again on the same device. In that case, give a CLOSE command to close the device, and then another SET LINE command to re-open the same device. Or rebuild your version of Kermit with the -DCLSOPN compile-time switch.
The DIAL command says "To cancel: Type your interrupt character (normally Ctrl-C)." This is just one example of where program messages and documentation assume your interrupt character is Ctrl-C. But it might be something else. In most (but not necessarily all) cases, the character referred to is the one that generates the SIGINT signal. If Ctrl-C doesn't act as an interrupt character for you, type the Unix command "stty -a" or "stty all" or "stty everything" to see what your interrupt character is. (Kermit could be made to find out what the interrupt character is, but this would require a lot of platform-dependent coding and #ifdefs, and a new routine and interface between the platform-dependent and platform-independent parts of the program.)
In general, the hangup operation on a serial communication device is prone to failure. C-Kermit tries to support many, many different kinds of computers, and there seems to be no portable method for hanging up a modem connection (i.e. turning off the RS-232 DTR signal and then turning it back on again). If HANGUP, DIAL, and/or Ctrl-\H do not work for you, and you are a programmer, look at the tthang() function in ckutio.c and see if you can add code to make it work correctly for your system, and send the code to the address above. (NOTE: This problem has been largely sidestepped as of edit 188, in which Kermit first attempts to hang up the modem by "escaping back" via +++ and then giving the modem's hangup command, e.g. ATH0, when DIAL MODEM-HANGUP is ON, which is the default setting.)
Even when Kermit's modem-control software is configured correctly for your computer, it can only work right if your modem is also configured to assert the CD signal when it is connected to the remote modem and to hang up the connection when your computer drops the DTR signal. So before deciding Kermit doesn't work with your modem, check your modem configuration AND the cable (if any) connecting your modem to the computer -- it should be a straight-through modem cable conducting the signals FG, SG, TD, RD, RTS, CTS, DSR, DTR, CD, and RI.
Many Unix systems keep aliases for dialout devices; for example, /dev/acu might be an alias for /dev/tty00. But most of these Unix systems also use UUCP lockfile conventions that do not take this aliasing into account, so if one user assigns (e.g.) /dev/acu, then another user can still assign the same device by referring to its other name. This is not a Kermit problem -- Kermit must follow the lockfile conventions used by the vendor-supplied software (cu, tip, uucp).
The SET FLOW-CONTROL KEEP option should be given *before* any communication (dialing, terminal emulation, file transfer, INPUT/OUTPUT/TRANSMIT, etc) is attempted, if you want C-Kermit to use all of the device's preexisting flow-control related settings. The default flow-control setting is XON/XOFF, and it will take effect when the first communication-related command is given, and a subsequent SET FLOW KEEP command will not necessarily know how to restore *all* of the device's original flow-control settings.
If file transfer does not work through a host to which you have rlogin'd, use "rlogin -8" rather than "rlogin". If that doesn't work, tell both Kermit programs to "set parity space".
The Encore TELNET server does not allow long bursts of input. When you have a TELNET connection to an Encore, tell C-Kermit on the Encore to SET RECEIVE PACKET-LENGTH 200 or thereabouts.
SET FLOW RTS/CTS is available in Unix C-Kermit only when the underlying operating system provides an Application Program Interface (API) for turning this feature on and off under program control, which turns out to be a rather rare feature among Unix systems. To see if your Unix C-Kermit version supports hardware flow control, type "set flow ?" at the C-Kermit prompt, and look for "rts/cts" among the options. Other common situations include:
System V R4 based Unixes are supposed to supply a <termiox.h> file, which gives Kermit the necessary interface to command the terminal driver to enable/disable hardware flow control. Unfortunately, but predictably, many implementations of SVR4 whimsically place this file in /usr/include/sys rather than /usr/include (where SVID clearly specifies it should be; see SVID, Third Edition, V1, termiox(BA_DEV). Thus if you build C-Kermit with any of the makefile entries that contain -DTERMIOX or -DSTERMIOX (the latter to select <sys/termiox.h>), C-Kermit will have "set flow rts/cts" and possibly other hardware flow-control related commands. BUT... That does not necessarily mean that they will work. In some cases, the underlying functions are simply not coded into the operating system.
WARNING: When hardware flow control is available, and you enable in Kermit on a device that is not receiving the CTS signal, Kermit can hang waiting for CTS to come up. This is most easily seen when the local serial port has nothing plugged in to it, or is connected to an external modem that is powered off.
C-Kermit is not a terminal emulator. Refer to page 147 of Using C-Kermit, 2nd Edition: "Most versions of C-Kermit -- Unix, VMS, AOS/VS, VOS, etc -- provide terminal connection without emulation. These versions act as a 'semitransparent pipe' between the remote computer and your terminal, terminal emulator, console driver, or window, which in turn emulates (or is) a specific kind of terminal." The environment in which you run C-Kermit is up to you.
If you are an X Windows user, you should be aware of an alternative to xterm that supports VT220 emulation, from Thomas E. Dickey:
http://dickey.his.com/xterm/xterm.html
Unix C-Kermit's SET KEY command currently can not be used with keys that generate "wide" scan codes or multibyte sequences, such as workstation function or arrow keys, because Unix C-Kermit does not have direct access to the keyboard.
However, many Unix workstations and/or console drivers provide their own key mapping feature. With xterm, for example, you can use 'xmodmap' ("man xmodmap" for details); here is an xterm mapping to map the Sun keyboard to DEC VT200 values for use with VT-terminal oriented applications like VMS EVE:
keycode 101=KP_0 keycode 119=KP_1 keycode 120=KP_2 keycode 121=KP_3 keycode 98=KP_4 keycode 99=KP_5 keycode 100=KP_6 keycode 75=KP_7 keycode 76=KP_8 keycode 77=KP_9 keycode 52=KP_F1 keycode 53=KP_F2 keycode 54=KP_F3 keycode 57=KP_Decimal keycode 28=Left keycode 29=Right keycode 30=KP_Separator keycode 105=KP_F4 keycode 78=KP_Subtract keycode 8=Left keycode 10=Right keycode 32=Up keycode 33=Down keycode 97=KP_Enter
Users of Linux consoles can use loadkeys ("man dumpkeys loadkeys keytables" for details. The format used by loadkeys is compatible with that used by Xmodmap, although it is not definitely certain that the keycodes are compatible for different keyboard types (e.g. Sun vs HP vs PC, etc).
If uploads (or downloads) fail immediately, give the CAUTIOUS command to Kermit and try again. If they still fail, then try SET PREFIXING ALL. If they still fail, try SET PARITY SPACE. If they still fail, try ROBUST.
If uploads (particularly of large files and/or binary files) begin successfully but then fail constently after a certain amount of bytes have been sent, check:
If none of these seem to explain it, then the problem is not size related, but reflects some clash between the file contents and the characteristics of the connection, in which case follow the instructions in the first paragraph of this section.
Suppose two copies of Kermit are receiving files into the same directory, and the files have the same name, e.g. "foo.bar". Whichever one starts first opens an output file called "foo.bar". The second one sees there is already a foo.bar file, and so renames the existing foo.bar to foo.bar.~1~ (or whatever). When the first file has been received completely, Kermit goes to change its modification time and permissions to those given by the file sender in the Attribute packet. But in Unix, the APIs for doing this take a filename, not a file descriptor. Since the first Kermit's file has been renamed, and the second Kermit is using the original name, the first Kermit changes the modtime and permissions of the second Kermit's file, not its own. Although there might be a way to work around this in the code, e.g. using inode numbers to keep track of which file is which, this would be tricky and most likely not very portable. It's better to set up your application to prevent such things from happening, which is easy enough using the script language, filename templates, etc.
Suppose you start C-Kermit with a command-line argument to send or receive a file (e.g. "kermit -r") and then type Ctrl-\c immediately afterwards to escape back and initiate the other end of the transfer, BUT your local Kermit's escape character is not Ctrl-\. In this case, the local Kermit passes the Ctrl-\ to the remote system, and if this is Unix, Ctrl-\ is likely to be its SIGQUIT character, which causes the current program to halt and dump core. Well, just about the first thing C-Kermit does when it starts is to disable the SIGQUIT signal. However, it is still possible for SIGQUIT to cause Kermit to quit and dump core if it is delivered while Kermit is being loaded or started, before the signal can be disabled. There's nothing Kermit itself can do about this, but you can prevent it from happening by disabling SIGQUIT in your Unix session. The command is usually something like:
stty quit undef
Unix C-Kermit does not reject incoming files on the basis of size. There appears to be no good (reliable, portable) way to determine in advance how much disk space is available, either on the device, or (when quotas or other limits are involved) to the user.
Unix C-Kermit discards all carriage returns from incoming files when in text mode.
If C-Kermit has problems creating files in writable directories when it is installed setuid or setgid on BSD-based versions of Unix such as NeXTSTEP 3.0, it probably needs to be rebuilt with the -DSW_ACC_ID compilation switch.
If you SET FILE DISPLAY FULLSCREEN, and C-Kermit complains "Sorry, terminal type not supported", it means that the terminal library (termcap or termlib) that C-Kermit was built with does not know about a terminal whose name is the current value of your TERM environment variable. If this happens, but you want to have the fullscreen file transfer display, EXIT from C-Kermit and set a Unix terminal type from among the supported values that is also supported by your terminal emulator, or else have an entry for your terminal type added to the system termcap and/or terminfo database.
If you attempt to suspend C-Kermit during local-mode file transfer and then continue it in the background (via bg), it will block for "tty output" if you are using the FULLSCREEN file transfer display. This is apparently a problem with curses. Moving a local-mode file transfer back and forth between foreground and background works correctly, however, with the SERIAL, CRT, BRIEF, or NONE file transfer displays.
If C-Kermit's command parser no longer echoes, or otherwise acts strangely, after returning from a file transfer with the fullscreen (curses) display, and the curses library for your version of Unix includes the newterm() function, then try rebuilding your version of C-Kermit with -DCK_NEWTERM. Similarly if it echoes doubly, which might even happen during a subsequent CONNECT session. If rebuilding with -DCK_NEWTERM doesn't fix it, then there is something very strange about your system's curses library, and you should probably not use it. Tell C-Kermit to SET FILE DISPLAY CRT, BRIEF, or anything else other than FULLSCREEN, and/or rebuild without -DCK_CURSES, and without linking with (termlib and) curses. Note: This problem seemed to have escalated in C-Kermit 7.0, and -DCK_NEWTERM had to be added to many builds that previously worked without it: Linux, AIX 4.1, DG/UX, etc. In the Linux case, it is obviously because of changes in the (n)curses library; the cause in the other cases is not known.
C-Kermit creates backup-file names (such as "oofa.txt.~1~") based on its knowledge of the maximum filename length on the platform where it is running, which is learned at compile time, based on MAXNAMLEN or equivalent symbols from the system header files. But suppose C-Kermit is receiving files on a Unix platform that supports long filenames, but the incoming files are being stored on an NFS-mounted file system that supports only short names. NFS maps the external system to the local APIs, so C-Kermit has no way of knowing that long names will be truncated. Or that C-Kermit is running on a version of Unix that supports both long-name and short-name file systems simultaneously (such as HP-UX 7.00). This can cause unexpected behavior when creating backup files, or worse. For example, you are sending a group of files whose names are differentiated only by characters past the point at which they would be truncated, each file will overwrite the previous one upon arrival.
SECTION CONTENTS
11.1. C-Kermit as an External Protocol 11.2. Invoking External Protocols from C-Kermit
Unix C-Kermit can be used in conjunction with other communications software in various ways. C-Kermit can be invoked from another communications program as an "external protocol", and C-Kermit can also invoke other communication software to perform external protocols.
This sort of operation makes sense only when you are dialing out from your Unix system (or making a network connection from it). If the Unix system is the one you have dialed in to, you don't need any of these tricks. Just run the desired software on your Unix system instead of Kermit. When dialing out from a Unix system, the difficulty is getting two programs to share the same communication device in spite of the Unix UUCP lockfile mechanism, which would normally prevent any sharing, and preventing the external protocol from closing (and therefore hanging up) the device when it exits back to the program that invoked it.
(This section deleted; see Using C-Kermit, 2nd Ed, Chapter 14.)
"pcomm" is a general-purpose terminal program that provides file transfer capabilities itself (X- and YMODEM variations) and the ability to call on external programs to do file transfers (ZMODEM and Kermit, for example). You can tell pcomm the command to send or receive a file with an external protocol:
Send Receive
ZMODEM sz filename rz
Kermit kermit -s filename kermit -r
pcomm runs external programs for file transfer by making stdin and stdout point to the modem port, and then exec-ing "/bin/sh -c xxx" (where xxx is the appropriate command). However, C-Kermit does not treat stdin and stdout as the communication device unless you instruct it:
Send Receive
Kermit kermit -l 0 -s filename kermit -l 0 -r
The "-l 0" option means to use file descriptor 0 for the communication device.
In general, any program can pass any open file descriptor to C-Kermit for the communication device in the "-l" command-line option. When Kermit is given a number as the argument to the "-l" option, it simply uses it as a file descriptor, and it does not attempt to close it upon exit.
Here's another example, for Seyon (a Linux communication program). First try the technique above. If that works, fine; otherwise... If Seyon does not give you a way to access and pass along the file descriptor, but it starts up the Kermit program with its standard i/o redirected to its (Seyon's) communications file descriptor, you can also experiment with the following method, which worked here in brief tests on SunOS. Instead of having Seyon use "kermit -r" or "kermit -s filename" as its Kermit protocol commands, use something like this (examples assume C-Kermit 6.0):
kermit -YqQl 0 -r <-- to receive kermit -YqQl 0 -s filename(s) <-- to send one or more files
kermit -YqQF 0 -r <-- to receive kermit -YqQF 0 -s filename(s) <-- to send one or more files
Command line options:
Y - skip executing the init file Q - use fast file transfer settings (default in 8.0) l 0 - transfer files using file descriptor 0 for a serial connection F 0 - transfer files using file descriptor 0 for a Telnet connection q - quiet - no messages r - receive s - send
[ Top ] [ Contents ] [ Section Contents ] [ Previous ]
(This section is obsolete, but not totally useless. See Chapter 14 of Using C-Kermit, 2nd Edition).
After you have opened a communication link with C-Kermit's SET LINE (SET PORT) or SET HOST (TELNET) command, C-Kermit makes its file descriptor available to you in the \v(ttyfd) variable so you can pass it along to other programs that you RUN from C-Kermit. Here, for example, C-Kermit runs itself as an external protocol:
C-Kermit>set modem type hayes C-Kermit>set line /dev/acu C-Kermit>set speed 2400 C-Kermit>dial 7654321 Call complete. C-Kermit>echo \v(ttyfd) 3 C-Kermit>run kermit -l \v(ttyfd)
Other programs that accept open file descriptors on the command line can be started in the same way.
You can also use your shell's i/o redirection facilities to assign C-Kermit's open file descriptor (ttyfd) to stdin or stdout. For example, old versions of the Unix ZMODEM programs, sz and rz, when invoked as external protocols, expect to find the communication device assigned to stdin and stdout with no option for specifying any other file descriptor on the sz or rz command line. However, you can still invoke sz and rz as exterior protocols from C-Kermit if your current shell ($SHELL variable) is ksh (the Korn shell) or bash (the Bourne-Again shell), which allows assignment of arbitrary file descriptors to stdin and stdout:
C-Kermit> run rz <&\v(ttyfd) >&\v(ttyfd)
or:
C-Kermit> run sz oofa.zip <&\v(ttyfd) >&\v(ttyfd)
In version 5A(190) and later, you can use C-Kermit's REDIRECT command, if it is available in your version of C-Kermit, to accomplish the same thing without going through the shell:
C-Kermit> redirect rz
or:
C-Kermit> redirect sz oofa.zip
A complete set of rz,sz,rb,sb,rx,sx macros for Unix C-Kermit is defined in the file ckurzsz.ini. It automatically chooses the best redirection method (but is redundant since C-Kermit 6.0, which now has built-in support for external protocols via its SET PROTOCOL command).
Note that external protocols can be used on C-Kermit SET LINE or SET HOST connections only if they operate through standard input and standard output. If they open their own connections, Kermit can't redirect them over its own connection.
As of version 7.0, C-Kermit supports a wide range of security options for authentication and encryption: Kerberos 4, Kerberos 5 / GSSAPI, SSL/TLS, and SRP. See the separate security document for details.
Date: Thu, 12 Mar 92 1:59:25 MEZ
From: Walter Mecky <walter@rent-a-guru.de>
Subject: Help.Unix.sw
To: svr4@pcsbst.pcs.com, source@usl.com
PRODUCT: Unix
RELEASE: Dell SVR4 V2.1 (is USL V3.0)
MACHINE: AT-386
PATHNAME: /usr/lib/libc.so.1
/usr/ccs/lib/libc.a
ABSTRACT: Function ttyname() does not close its file descriptor
DESCRIPTION:
ttyname(3C) opens /dev but never closes it. So if it is called
often enough the open(2) in ttyname() fails. Because the broken
ttyname() is in the shared lib too all programs using it can
fail if they call it often enough. One important program is
uucico which calls ttyname for every file it transfers.
Here is a little test program if your system has the bug:
#include <stdlib.h>
#include <stdio.h>
main() {
int i = 0;
while (ttyname(0) != NULL)
i++;
perror("ttyname");
printf("i=%d\n", i);
}
If this program runs longer than some seconds you don't have the bug.
WORKAROUND: None FIX: Very easy if you have source code.
Another user reports some more explicit symptoms and recoveries:
> What happens is when invoking ckermit we get one of the following > error messages: > You must set line > Not a tty > No more processes. > One of the following three actions clears the peoblem: > shutdown -y -g0 -i6 > kill -9 the ttymon with the highest PID > Invoke sysadm and disable then enable the line you want to use. > Turning off respawn of sac -t 300 and going to getty's and uugetty's > does not help. > > Also C-Kermit reports "?timed out closing /dev/ttyxx". > If this happens all is well. ------------------------------(Note: the following problem also occurs on SGI and probably many other Unix systems):
From: James Spath <spath@jhunix.hcf.jhu.edu>
To: Info-Kermit-Request@cunixf.cc.columbia.edu
Date: Wed, 9 Sep 1992 20:20:28 -0400
Subject: C-Kermit vs uugetty (or init) on Sperry 5000
We have successfully compiled the above release on a Unisys/Sperry 5000/95. We used the sys5r3 option, rather than sys5r2 since we have VR3 running on our system. In order to allow dialout access to non-superusers, we had to do "chmod 666 /dev/tty###, where it had been -rw--w--w- (owned by uucp), and to do "chmod +w /usr/spool/locks". We have done text and binary file transfers through local and remote connections.
The problem concerning uucp ownership and permissions is worse than I thought at first. Apparently init or uugetty changes the file permissions after each session. So I wrote the following C program to open a set of requested tty lines. I run this for any required outgoing line prior to a Kermit session.
------ cut here -------
/* opentty.c -- force allow read on tty lines for modem i/o */
/* idea from: restrict.c -- System 5 Admin book Thomas/Farrow p. 605 */
/* /jes jim spath {spath@jhunix.hcj.jhu.edu } */
/* 08-Sep-92 NO COPYRIGHT. */
/* this must be suid to open other tty lines */
/* #define DEBUG */
#define TTY "/dev/tty"
#define LOK "/usr/spool/locks/LCK..tty"
#include <stdio.h>
/* allowable lines: */
#define TOTAL_LINES 3
static char allowable[TOTAL_LINES][4] = { "200", "201", "300" };
static int total=TOTAL_LINES;
int allow;
/* states: */
#define TTY_UNDEF 0
#define TTY_LOCK 1
#define TTY_OKAY 2
main(argc, argv)
int argc; char *argv[]; {
char device[512];
char lockdev[512];
int i;
if (argc == 1) {
fprintf(stderr, "usage: open 200 [...]\n");
}
while (--argc > 0 && (*++argv) != NULL ) {
#ifdef DEBUG
fprintf(stderr, "TRYING: %s%s\n", TTY, *argv);
#endif
sprintf(device, "%s%s", TTY, *argv);
sprintf(lockdev, "%s%s", LOK, *argv);
allow = TTY_UNDEF; i = 0;
while (i <= total) { /* look at all defined lines */
#ifdef DEBUG
fprintf(stderr, "LOCKFILE? %s?\n", lockdev);
#endif
if (access(lockdev, 00) == 0) {
allow=TTY_LOCK;
break;
}
#ifdef DEBUG
fprintf(stderr, "DOES:%s==%s?\n", allowable[i], *argv);
#endif
if (strcmp(allowable[i], *argv) == 0)
allow=TTY_OKAY;
i++;
}
#ifdef DEBUG
fprintf(stderr, "allow=%d\n", allow);
#endif
switch (allow) {
case TTY_UNDEF:
fprintf (stderr, "open: not allowed on %s\n", *argv);
break;
case TTY_LOCK:
fprintf (stderr, "open: device locked: %s\n", lockdev);
break;
case TTY_OKAY:
/* attempt to change mode on device */
if (chmod (device, 00666) < 0)
fprintf (stderr, "open: cannot chmod on %s\n", device);
break;
default:
fprintf (stderr, "open: FAULT\n");
}
}
exit (0);
}
Unix versions, especially those for PCs (SCO, Unixware, etc) might be augmented by third-party communication-board drivers from Digiboard, Stallion, etc. These can sometimes complicate matters for Kermit considerably since Kermit has no way of knowing that it is going through a possibly nonstandard driver. Various examples are listed in the earlier sections of this document; search for Stallion, Digiboard, etc. Additionally:
- World Wide Escalation Support, Stallion Technologies, Toowong QLD, support@stallion.oz.au.
Later (December 1997, from the same source):
[ Top ] [ Contents ] [ C-Kermit Home ] [ C-Kermit 8.0 Overview ] [ Kermit Home ]