C-Kermit for Windows DOWNLOAD LINKS; Number 1 is the choice for most people. x86 and ARM32 are32 bits; the others are 64 bits. CLICK HERE for more information about these builds.

Beta.09 sections: BSD  HPUX  Linux  Custom  Mac  Sun  VMS 

Footnotes:

1. The HP92453-01 A.10.32.00 Bundled compiler, as opposed to the premium one that supports ANSI C and costs extra.
2. Full make command: LD_LIBRARY_PATH=/usr/local/lib/hpux32/ gmake MAKE=gmake 'KSSLINC=-I/usr/local/include' 'KSSLLIB=-L/usr/local/lib/hpux32/' hpux1100gcc+openssl
3. Full make command: gmake hpux1100gcc MAKE=gmake KFLAGS='-mlp64 LNKFLAGS='-mlp64 -L /usr/local/lib/hpux64'
4. Full make command: gmake MAKE=gmake KFLAGS='-mlp64' LNKFLAGS='-mlp64' 'KSSLINC=-I/usr/local/include' 'KSSLLIB=-L/usr/local/lib/hpux64/' hpux1100gcc+openssl
5. Full make command: gmake MAKE=gmake KFLAGS='-mlp64' LNKFLAGS='-mlp64' KZLIBDIR='/opt/zlib/lib' \
   'KSSLINC=-I/usr/local/include' 'KSSLLIB=-L/usr/local/lib/hpux64/' hpux1100gcc+openssl
6. Full make command: LD_LIBRARY_PATH=/opt/openssl3/lib/hpux64/ gmake MAKE=gmake KFLAGS='-mlp64' LNKFLAGS='-mlp64' 'KSSLINC=-I/opt/openssl3/include' \
   'KSSLLIB=-L/opt/openssl3/lib/hpux64' hpux1100gcc+openssl
7. It seems the timeb.h library along with the functions it declared, such as ftime(), are now "deprecated" and "should not be used". Of course the Kermit code was written long before this edict was issued, and not only that, but C-Kermit must continue to compile and run on older platforms that depend on timeb.h and its functions. In the meantime, builds that fail with errors like "ckufio.c:211:10: fatal error: 'sys/timeb.h' file not found" can be repaired by "KFLAGS=-DNOSYSTIMEBH" as in the OpenBSD 7.2 build.

0. The first two HP/UX B.10.20 builds were done by me on an HP 9000/715 with the bundled non-ANSI compiler; KFLAGS=-DNOIKSD was required to block an #include for <varargs.h>, which was AWOL, plus a patch to ckutio.c to put the #include within #ifndef NOIKSD..#endif. All the other HP B.10.20 listings are from PeterE on an HP 9000/785. HP/UX 10.20 dates from 1996 so it's a pretty good testbed for old compilers, headers, and libs. After the failed build, an appropriate copy of <varargs.h> was found and installed.
1. 40 warnings because Clang wants unnecessary brackets in IF conditions that include && and ||. Compilation still completes and the well-defined and time-honored precedence of && and !! is honored in execution. There is a greater risk in "fixing" all these these perfectly legal constructions by hand than in leaving them as they are.
2. ckuus2.c dohfunc(): 3 instances of "warning: string length 'nnn' is greater than the length '509' ISO C90 compilers are required to support" (but they are accepted anyway).
3. ckufio.c makes six syslog() calls between lines 1406 and 4374, that use the %m gnu_printf format, which sets off alarm bells. As of Beta.07, syslog() is no longer included in C-Kermit builds unless you ask for it explicitly, e.g. "make linux KFLAGS=-DDOSYSLOG".
4. In ckcrp.c: implicit declarations of des_random_key, des_random_seed, des_ecb_encrypt, des_set_odd_parity, des_key_sched. In ck_ssl.c: des_ecb3_encrypt.
5. Sun Studio 12.6 with cc patch 152881-06. Solaris 11.4 is from 2018.
6. Debian builds were all on the development version of Debian (the next release after 11.5); development versions do not have a version number. they were all done with OpenSSL 3.0.5-4. Every build had these two warnings because C-Kermit deliberately records the date and time it was built:
   • ckuus5.c:12669: __DATE__ and __TIME__ "might prevent reproducible builds"
     (one warning each for __DATE__ and __TIME__, only in builds with 27 warnings)
   • In all builds, ck_ssl.c: 25 "deprecated" warnings about OpenSSL APIs.
7. Debian builds with OpenSSL 3.0.5-4 (these are fixed in Beta.07):
   • In 9 builds, ckuusr.c:8118:24: warning: '%s' directive output may be truncated writing up to 2147483635 bytes into a region of size 319
   • ckudia.c:6602:48: warning: '%s' directive writing up to 51 bytes into a region of size 50
8. SunOS 4.1.1 dates from March 1990, and the compiler (which has no version number) is 100% non-ANSI, an excellent test of C-Kermit's backwards compatilibility. C-Kermit code produced 3 warnings:
   "ckucmd.c", line 8205: warning: statement not reached
   "ckuus5.c", line 1267: warning: illegal pointer combination
   "ckuus5.c", line 1268: warning: illegal pointer combination.
   Meanwhile ckutio.c includes both <termios.h> and <sys/ioctl.h> and these two header files each define the same symbols: ECHO, NL0, NL1, TAB0, etc etc (19 of them), resulting in 19 "xxx redefined" warnings. Aside from that the build was smooth, but it took 100 minutes.
9. These builds require changes to ckcdeb.h, ckcmai.c, ckuus4.c, ckuus5.c, ckuus6.c, ckuusy.c, ckuusx.c, and the makefile, which will be in Beta.07.
10. The builder (SMS) had to build a 64-bit version of zlib, and manually install its products in "/usr/local/lib/hpux64/" for the HPUX11 SSL build to work.
11. netbsd+ssl and hpux1100gcc+openssl with OpenSSL 3.0.7 both reported references to the following symbols which are "deprecated" (number of references in parens): DH_free (5), DH_new (5), DH_set0_pqg (5), PEM_read_bio_DHparams (1), RSA_free (3), RSA_generate_key_ex (2), RSA_new (2), SSL_CTX_set_tmp_dh_callback (2), TLSv1_server_method (1).
12. From SMS: "For some still-unknown reason, this new OpenSSL installation seems imperfect:
    dyi# export PATH="/opt/openssl3/bin:$PATH"
    dyi# openssl version
    /usr/lib/hpux64/dld.so: Unable to find library 'libssl.so.3'.
    Killed
    Manually pointing to the directory works around the problem:
    dyi# LD_LIBRARY_PATH=/opt/openssl3/lib/hpux64/ openssl version
    OpenSSL 3.0.7 1 Nov 2022 (Library: OpenSSL 3.0.7 1 Nov 2022)
    This could be my fault or an OpenSSL defect, but presumably if the OpenSSL stuff were right, then the extra LD_LIBRARY_PATH prefix on the following build commands would be unnecessary." Note: I had the exact same problem on NetBSD 9.3 with OpenSSL 3.0.7.
13. Linux Fedora make linux+ssl warnings:
    ckcftp.c: incompatible pointer types: &gss_mech_krb5.
    ckuath.c: type defaults to int in declaration: tcp_rdns.
    ckuath.c: implicit function declarations: tls_is_anon, krb5_auth_con_getlocalsubkey, krb5_auth_con_getrecvsubkey, locate_txt_rr, krb5_get_init_creds_password, decode_krb5_ticket, krb5_read_message, krb5_write_message, k5_u2u_read_msg, k5_u2u_write_msg, net_read, net_write, rlog_ctrl.  ck_ssl.c: deprecated: RSA_generate_key_ex, DH_new, DH_free, DH_set0_pqg, PEM_read_bio_DHparams, dls_ctx, SSL_CTX_set_tmp_dh_callback.
14. The "normal clang warnings" are about "*s2++;" (see just below) and about complex IF statements with && (logical AND) and || (logical OR) not having parentheses, which are (of course) not required because the precedence of these operators is and always has been fully spelled out in the specifications and standards (see Note 1 above).

About *s2++;

Beta.05

Footnotes:

1. "*s2++;" expression result unused (see notes).
    ckuus5.c:6840 dangling else: "} else {" — doesn't look danging to me!
2. This one I don't understand:
   

   ckutio.c:4951:30: warning: ISO C does not allow extra ';' outside of a function [-Wpedantic]
   4951 | char * uucplockdir = LOCK_DIR;

   The other six are "ISO C does not support the '%m' gnu_printf format", which occur in the "deprecated" syslog code. If I build with -DNOSYSLOG or -DNODEPRECATED these warnings don't appear.
3. ckuus2.c dohfunc(): 3 instances of "warning: string length 'nnn' is greater than the length '509' ISO C90 compilers are required to support. (but they are accepted anyway).
4. 1 warning for the "*s2++; and 18 warnings because it wants IF conditions containing && or || to be enclosed in parentheses. "Pedantic" is right! The C language has always had clear and explicit rules about the order of precedence of these operators.
5. If I include an -I clause in KFLAGS to let it know where profile.h is, then it can't find com_err.h. If I do the same for that one, then it gets hundreds and hundreds of errors in ckuath.c: deprecated; structure has no member with that name; subscripted value is not a pointer, an array, or a vector; incompatible type for argument, and on and on.
6. HP-UX builds by Peter Eichhorn. At first all the builds failed. The following change was required:

   In ckuusr.c at line 12345, see this code:

   #ifdef TYPEINTERPRET
           type_intrp = 0;                 /* Always start with this off */
   #endif  /* TYPEINTERPRET */
   

   Move it down about 13 lines, below "#endif /* UNICODE */".

   This assignment statement is in the middle of a variable declaration list in the docmd() function. This should be an error everywhere but the code compiled cleanly on 25 different builds on other platforms, some of them with the strictest modern C compilers including clang and gcc -pedantic. So (at least so far) it's not serious enough to warrant another Beta test. If anybody gets the same error on any other platform, please let me know.
7. All Debian builds done against OpenSSL 3.0.5-4. All of them had warnings 1 and 2, the ones with 37 warnings also had 3 and 4:
   1. ckuus5.c:12609: macro "__DATE__" might prevent reproducible builds
   2. ckssl.c:850: 35 × [item] is deprecated: Since OpenSSL 3.0
   3. ckudia.c:6602:48: warning: '%s' directive writing up to 31 bytes into a region of size 30
   4. ckuusr.c:8118:24: warning: '%s' directive output may be truncated writing up to 2147483635 bytes into a region of size 319

Beta.04

• The first (Open)VMS C-Kermit build on x86_64 PC architecture... ever!
  
• The first C-Kermit for Windows (formerly known as Kermit 95) since 2003.

Notes:

1. The warnings will be looked at in a subsequent Beta test.
2. -DNOARROWKEYS will be necessary when Gnu kills the API used for reading them.
3. 87 warnings, then compilation stopped in ckcnet.c with errors in the rlog_* variables (in support code for rlogin protocol). This didn't happen in "make linuxclang", but that's much older clang version.
4. Linux-script-only: A minimum-size version for Linux that does only scripting and serial communication -- no networks, no file transfer, no security.
5. Type cast needed in ckctel.c to suppress the following warning:
   

       Compiling WISP$DUA0:[UTILITY.SOURCE.KERMIT.CKU10_0-BETA04]CKCTEL.C
                             sprintf(request,"%cUSER",TEL_ENV_VAR); /* safe */
           ..........................^
   

   %CC-W-PTRMISMATCH1, In this statement, the referenced type of the pointer value "request" is "unsigned char", which is not compatible with "char" because they differ by signed/unsigned attribute. At line number 4676 in WISP$DUA0:[UTILITY.SOURCE.KERMIT.CKU10_0-BETA04]CKCTEL.C;1.
6. All VMS builds with vendor-supplied compilers get an info message like the following:

       Compiling WISP$DUA0:[UTILITY.SOURCE.KERMIT.CKU10_0-BETA04]CKVTIO.C
                   while ((n--) && xx_inc(2) >= 0) ; /* Ignore Warning - see comments */
           ........................^
   

   %CC-I-QUESTCOMPARE, In this statement, the unsigned expression "(--ttxbn>=0?(unsigned)(ttxbuf[ttxbp++]&0X000000ff):txbufr(...))" is being compared with a relational operator to a constant whose value is not greater than zero. This might not be what you intended. At line number 921 in WISP$DUA0:[UTILITY.SOURCE.KERMIT.CKU10_0-BETA04]CKVTIO.C;1. %VCG-I-SUMMARY, Completed with 0 error(s), 0 warning(s), and 1 informational messages. At line number 4305 in WISP$DUA0:[UTILITY.SOURCE.KERMIT.CKU10_0-BETA04]CKCTEL.C;1.
7. 69 dangling else (suppressible by adding "-Wdangling-else"), 4 CHAR/char complaints (suppressible with "-funsigned-char" or "-Wpointer-sign"). 2 comparisons to string literals (suppressible with "-Wstring-compare"), 2 "logwtmp is deprecated" (suppressible with "-Wdeprecated-declarations" or with "-DNOWTMP"), 1 unused value.
8. Solaris 11 32-bit, compiled under Sun Studio 12.6, Sun C 5.15 SunOS_sparc 152881-06. ckuusx.3: 4 × "statement not reached" in tgetent, tgetstr, tputs, tgoto. ckufio.c: 3 × "statement not reached" at lines 1355, 6959, 7128. ckudia.c: 3 × "statement not reached" at lines 4551, 4586, 6917. ckuscr.c: 1 × "statement not reached" at line 157. ckcnet.c: 12 × argument incompatible with prototype at lines 615-623. ckctel.c: argument incompatible with prototype at line 255. ckcftp.c: 3 × argument incompatible with prototype at lines 615-622.
9. Solaris 11 64-bit, compiled under Sun Studio 12.6, Sun C 5.15 SunOS_sparc 152881-06. Warnings same as 32-bit build except there were no warnings in ckcnet.c.
10. By SMS, 14 July 2022. The first known builds for VMS C-Kermit for x86_64 architecture; cross-compiled on IA86 (a.k.a. Itanium). A few minor code changes were necessary which will show up in the next Beta, as will an updated VMS C-Kermit build procedure. Native builds should straightforward once the VMS x86_64 C compiler is released (any month now!). The SSL build was done by cross-compiling on IA64 and then (because the IA64 SSL library was in IA64 format) moving the object files to OpenVMSx86_64 and linking there.
11. by David Goodwin in New Zealand. Based on C-Kermit 10.0 Beta.04, plus his previous work in 2013. See this page for details. Can make SSH connections via external relay. This build can be openly distributed to all countries because it has no strong encryption code.
12. NOARROWKEYS was necessary for the Android because it no longer legal to peek into the FILE structure. So apparently it's a movement and the same will be true for Linux and other Unix-based OS's in short order. Therefore NOARROWKEYS will be the default configuration beginning in Beta.05.
13. These C-Kermit 10.0 Beta.04 builds are from https://www.freshports.org/comms/kermit/, where the Kermit makefile is invoked with this front end. When built with Clang 13 -Wall -Wextra there are 2090 warnings. With gcc 10 -Wall -Wextra, 1704 warnings. Builds without -Wall -Wextra are shown for gcc and clang in the two builds for which sizes are given. According to the package maintainer, future builds will be "make freebsd", not "make freebsd+ssl".
14. The OpenBSD patches will be in the next Beta. Warnings: 69 "dangling else", 2 "comparison against a string literal".
15. Builds for AIX with IBM by Jeff Johnson. These all failed for one reason or another; fixes will be in the next Beta.
    1. In ckufio.c at line 2486, "if errno" should have been "if (errno)", which was fatal. No other compiler caught this one. Also this (nonfatal) warning: ckcdeb.h line 596, /* #endif /* NETPTY */   ("/*" detected in comment.)
    2. SSL is IBM OpenSSL 1.0.2u 20 Dec 2019.
    3. In time.h included from ckufio.c line 225, 'const time_t *' wrong type (fatal). Plus various warnings.
    4. Numerous warnings (dangling else, incompatible pointer types, etc). "if errno" was fatal.
16. 4×"Option 'pipe' not supported"; "Infinite loop. Program might not stop"; 4×"Additional optimization may be attained by recompiling and specifying XMEM option with a value greater than 8192"; 2×"Result of comparison against a string literal is unspecified"; 6×"'TLSv1 server method' deprecated".
17. Note: icc is the Intel C Classic compiler. Various warnings: 10×"argument not compatible with parameter" (strlen); "10×'ck_tn_decrypting' declared implicitly; ck_crp.c: extern encrypt_debug_mode; (explicit type missing); plus about 20 "deprecated" warnings in ck_ssl.c. The Intel OneAPI compilation had tons of "dangling else" warnings.
18. Characters are not echoed at the command prompt until the Enter key is pressed. But the same source compiled and STATICALLY linked on Ubuntu runs correctly on Alpine.

Beta.03