C-Kermit 10.0 Beta.05, 26 September 2022

[dcantrell]

1. ckcdeb.h: Removed the arrow-key feature from all platforms. Support for the API that detected arrow-key strokes is disappearing from one libc after another, so I might as well do it now for everybody. This was done near the top of ckcdeb.h, which (despite its name) is C-Kermit primary header file. I removed -DNOARROWKEYS from all the makefile targets that include it because it happens automatically now. If your platform (e.g. perhaps NetBSD or VMS) supports the API and you want the feature back, build C-Kermit with KFLAGS=-DDOARROWKEYS.
2. ckcmai.c: Straightened out some confusion with the program herald. There are two routines: makever() in ckcmai.c, and herald() in ckuus5.c. makever() creates PART of the version string but doesn't print anything. herald() creates another part of the version string and then prints both parts and some other stuff, and it also defines the builtin (herald) variable. For a hard-to-explain reason, we needed to have the herald text defined BEFORE the initialization file was executed, but then not printed until AFTER the command-line arguments were processed so "kermit -H" (skip printing the herald) still worked. This required moving chunks of code from herald() to makever(). Modules changed: ckcmai.c, ckuus5.c, and ckuusy.c.
3. ckcnet.c: "make xxx -DNORLOGIN" generated tons of errors. All the rlogin-related functions needed to protected by #ifndef NORLOGIN..#endif. Lines 7180-7567: rlog_ini(), rlog_ctrl(), rlog_oob(), rlogoobh().
4. ckuusx.c line 96: added prototypes for ck_tn_encrypting() and ck_tn_decrypting().
5. ck_ssl.c: add "return(0);" outside the #ifdefs to squelch "non-void function does not return a value" Clang warning.
6. ckcnet.c: in http_security(), cipher = SSL_get_current_cipher(tls_http_con); needed to be: cipher = (SSL_CIPHER *)SSL_get_current_cipher(tls_http_con);
7. ckuath.c: Added prototypes for encrypt_is_decrypting(), encrypt_request_end(), and get_crypt_table().
8. ck_crp.c: about line 3876: removed "extern int encrypt_debug_mode;". It is not extern and it was already declared above.

With changes 4-8 listed above, SSL builds generate no errors or warnings in OpenSSL 1.1.1d-fips 11 Feb 2013 (Red Hat 6.1) nor in OpenSSL 1.1.1f 11 Feb 2020 (Ubuntu 20.4.1). A few warnings are generated in OpenSSL 1.1.f 31 Mar 2020 on NetBSD 9.2. In all three cases compilation and linking succeed and a C-Kermit executable is produced.

25 September 2022:

• ckufio.c line 2486 "} else if errno {" should have been "} else if ( errno ) {" (fatal in AIX 15).
• ckufio.c line 2474 "debug(F110,"zchki stat errmsg",ck_errstr(),""); final argument should have been 0.
• ckctel.c lines 4679, 4681, 6014, 6221: fixes to address Clang 10.0.1 complaints about signed vs unsigned char in Telnet code (but why does it complain at all if it was given the "-funsigned-char" directive?). This was also a problem in VMS.
• ckuusr.c: changed 'if (s == "")' to 'if (s == (char *)0)' to silence Clang warnings.
• ckuus3.c SHOW DIAL: suppressed "dangling else" warnings for Clang.
• ckuus4.c in SHOW X25, SHOW PAD, SHOW TCP, SHOW TELNET, SHOW NETBIOS, SHOW NETWORK, etc etc: suppressed 155 "dangling else" warnings for Clang.
• ckuus5.c at line 6915 is "} else {" and Clang insists this is a dangling else. I just don't see it. This was the only Clang warning in ckuus5.c.
• ckuus6.c, "fixed" 8 dangling else's (in quotes because "dangling else" was not a crime when C-Kermit was first written. The structure and symantics of the IF statement were clearly spelled out by Kernighan and Ritchie. What good is a programming language that changes out from under its installed base of software?) With all this C-Kermit compiles with gcc -Wpedantic with no errors or wanings and with Clang 10.0.1 without errors and almost without warnings (there is still this one but really, so what?).

24 September 2022:

• Merged David Goodwin's Windows changes from the past several months into my code, a large change involving 26 files: ckcasc.h, ckcdeb.h, ckcfn2.c, ckcfns.c, ckcftp.c, ckcker.h, ckcmai.c, ckcnet.c, ckcpro.w, ckctel.c, ckcuni.h, ckuath.c, ckucmd.c, ckudia.c, ckuscr.c, ckuus2.c, ckuus3.c, ckuus4.c, ckuus5.c, ckuus6.c, ckuus7.c, ckuusr.c, ckuusr.h, ckuusx.c, ckuusy.c, ckuver.h. The changes are listed on David's change log, which is here: http://www.ext.zx.net.nz/software/notes/kermit95/#updates.
• Arrow-key support... The C-Kermit command parser has a scrollback buffer, and at the C-Kermit prompt you could always (and still can) use Ctrl-P to scroll up (to recall previous commands) and Ctrl-N to scroll back down. In C-Kermit 8.0 (2004) I added the ability to do this also with with the Up-arrow and Down-arrow keys, as people might expect. This relied on a somewhat obscure API. A few years ago this API was "deprecated", at least in Glibc. Now it seems to have been totally removed in some Linux distributions, and as a consequence C-Kermit builds fail. So I have to remove this feature. Honestly, I have no idea how widespread the problem is, but I do know it's present in Linux (on some, but not all, distributions; for example I see no problems on Ubuntu 20.04.1) and in Android. So to start I'm removing it in Linux and Android: + Before removing it, builds OK on Ubuntu 20.04.3 and arrow keys work + make linux KFLAGS=-DNOARROWKEYS builds OK and arrow keys don't work + ckcdeb.h has arrow-key ifdefs + but so does ckucmd.h, why? o Removed these from ckucmd.h o Built normally ok, arrow keys work ok. o Built again with NOARROWKEYS, ok, and arrow keys don't work. o So that clause did not need to be in ckucmd.h, so I removed it. + Added -DNOARROWKEYS for Linux. + If you want to build for Linux with arrow-key support, do "make linux KFLAGS=-DOARROWKEYS". + Added -DNOARROWKEYS to android.mk.
• ALPINE LINUX HELP NEEDED: A user of Alpine Linux (musl libc) reports that this C-Kermit version, although it builds normally ("make linux") and without complaint, does not work right: command keystrokes are not echoed until the Enter key is pressed. But without -DNOARROWKEYS, it doesn't build at all because the "obscure API" isn't there. Can anybody with Alpine Linux lend a hand?
• Added comments to ckucmd.[ch] about its misnomer (it's really for all platforms, not just Unix).
• Added askmore() (more prompting) to COPY /TOSCREEN.
• SHOW FILE command: if download directory not defined, say "(your current directory)" instead of "(none)" (ckuusr.c).
• Checked what happens if I give CK a SET SEND PACKET 10 command and then send a file: it works perfectly - S and F packets are not truncated but data packets are the requested length.
• HELP FUNC EVALUATE... improved help text (ckuus4.c)
• (herald) was wrong, I fixed it, ckuus5.c: herald() routine.
• ckcfn2.c: logpkt() - added a comment explaining the packet-log format.
• Made (fullversion) = "10.0.400 Beta.05"
• Added discriptive comment heading to ckuus5.c: cmcvtdate()
• Made a timeless cvtdate() using negative format arguments.
• Adjusted cvtdate() to report date only if format arg is negative.
• Adjusted HELP FUNC CVTDATE accordingly.
• Corrected a glitch in HELP DATE text: ckuus2.c.
• Cleaned up NEWS command a bit but needs work: ckuus2.c.

19 September 2022:

• Added new options to the COPY command: /TOSCREEN and /INTERPRET. Used together they do what TYPE /INTERPRET does everywhere except Windows: it displays a file on the screen with all its Kermit backslash-escapes replaced by their values. The code is mainly in ckuus6.c docopy().
• I also fixed TYPE /INTERPRET so it clears the intepret bit after it finishes. Previously every subsequent TYPE command would act like it included an /INTERPRET switch.
• Not C-Kermit specifically but I wrote a new document that updates the old and scattered information about Kermit packets: types, formats, encodings, block checks, short vs long, etc, everything in one place: https://www.kermitproject.org/kpackets.html.

17 September 2022:

ckcfns.c sdata():

1. Even after the previous change, C-Kermit could send a long-format packet that was one byte too long; fixed in the obvious way.
2. In tracking down this problem I also noticed that if I gave a SET RECEIVE PACKET-LENGTH or SET SEND PACKET-LENGTH command, this too could result in overlong packets in some cases. Fixed now.

14 September 2022:

ckcfns.c sdata():

1. Reduced maximum send-packet data size (spsiz) for short packets from 94 to 90 to avoid edge cases which could cause the encoded packet length to be out of range, i.e. not a printable 7-bit US-ASCII character.

ckcfn2.c spack():

1. Added code to check whether Long Packet capability had been negotiated before deciding to build a long packet.

2. Corrected an off-by-two error in calculating whether the packet would be have to be in long format.

   Before the previous two changes, file transfers from C-Kermit 10.0 (and some of the earlier 9.0.305 developement versions) to E-Kermit would fail. With the changes, they succeed. Also:

ckcdeb.h:

1. Disabled TYPE /INTERPRET on Windows until we can figure out how to make it work.

ckcmai.c:

1. Changed the numeric version number from 1000000 to 1000400, where the last three digits reflect the monotonically increasing edit number.

23 August 2022:

• I added a new option to the TYPE command: /INTERPRET. "type /interpret somefile.txt" displays the lines of the file with all Kermit backslash-escapes (variable references, etc) interpreted, i.e. replaced by their value. So for example, if you have a file hello.txt:

Welcome to (herald): pathname(\%0) This message is coming to you from your C-Kermit initialization file: pathname(~/.kermrc)

You can tell C-Kermit to "type /interpret hello.txt" and the
result will be something like:

Welcome to C-Kermit 10.0 OPEN SOURCE: Beta.05, 23 Aug 2022: /net/u/1/f/fdc/beta05/wermit This message is coming to you from your C-Kermit initialization file: /net/u/1/f/fdc/.kermrc

• For the benefit of packagers who depended on the old version-number format, I reinstated C-Kermit's edit number, but not for public view. The new C-Kermit still announces itself as C-Kermit 10.0. To see the edit number use the new built-in variable: (fullversion):

(~/beta05/) C-Kermit>show variable fullversion (fullversion) = 10.0.400 Beta.05 (~/beta05/) C-Kermit>echo (fullversion) (fullversion) = 10.0.400 Beta.05 (~/beta05/) C-Kermit>

Furthermore, the filenames of the tar and zip archives will be cku400.*, not cku10.0.* as they have been since 10.0 Beta.01, for conformity with all previous releases when they were cku300, cku301, cku302, etc). Of course "400" could be "401" or any higher number as development progresses. Also the tar and zip archive filenames will include the edit number, as they did before the first C-Kermit 10.0 Beta test.

• A bug that prevented C-Kermit for Windows (CKW) from including the Beta-test in the program herald ((herald)) was corrected.

19 August 2022:

I removed all the code that supported special version numbers for (in chronological order) Mac Kermit (circa 1985), Kermit/2 (for OS/2), and Kermit 95 (for MS Windows). From now all C-Kermits will identify themselves in the same way, for example:

C-Kermit 10.0 OPEN SOURCE: Beta.05, 19 Aug 2022, for NetBSD 9.2 (64-bit)

17 August 2022:

The second Beta test of C-Kermit for Windows (formerly known as Kermit 95) was released. This work was done by David Goodwin in New Zealand, and any changes he had to make to the non-Windows-specific modules will be incorporated into Beta.05.

Meanwhile more Beta.04 builds on platforms where it was tested previously would be most welcome: macOS (including old versions), HP-UX, MINIX, Solaris versions before 11, etc, not to mention platforms where it has not yet been tested at all: FreeBSD, OpenBSD, AIX, plus any discontinued OS's that are still running somewhere: IRIX, Tru64/Digital Unix, 4.xBSD, NeXTSTEP, System V, DG/UX, SCO, QNX, SINIX, SunOS, Xenix, etc).

[jaltman]

Overall this looks good but I would prefer if the whitespace errors were corrected before submission.

git rebase --whitespace=fix

For authorship please use

git commit --author="Kermit Project kermit@kermitproject.org"

At least for Frank's code drops.

Thank you for setting the author date (git commit --date) to the timestamp of the tarball.

Frank often fails to clean out his work files before he creates a tarball. In beta.05 for example there are

• ckcnet-BEFORE-NONET-FIX.c
• ckuus7-beta02.c
• ckuus7-save.c

I remove these work files from the tree before I create the commit. Since we have a git repository prior versions can be extracted from that.

I have merged the commit with the above changes as 0c0c21f5bd29605a256360c749d699e11f105c18

[dcantrell]

Overall this looks good but I would prefer if the whitespace errors were corrected before submission.

git rebase --whitespace=fix

Noted.

For authorship please use

git commit --author="Kermit Project kermit@kermitproject.org"

I was pretty sure I changed that as well, but if I didn't then I forgot while also fixing the timestamp.

At least for Frank's code drops.

Thank you for setting the author date (git commit --date) to the timestamp of the tarball.

That just felt correct to do.

Frank often fails to clean out his work files before he creates a tarball. In beta.05 for example there are

* ckcnet-BEFORE-NONET-FIX.c

* ckuus7-beta02.c

* ckuus7-save.c

I remove these work files from the tree before I create the commit. Since we have a git repository prior versions can be extracted from that.

Noted, I will make sure to clean those up.

I have merged the commit with the above changes as 0c0c21f

Sounds good. I will prepare a PR for the next release after this one and continue forward.

Frank was also asking about my membership in both the organization and the project, he may reach out to you for some help.