Created attachment 224809 [details] proposed working patch to 3.6.2-1 with poudriere=ok, portlint=ok minor changes (trivial-features upgrades in endianess): cffi (https://github.com/cffi/cffi/compare/v0.23.0...v0.24.1) cl-unicode cl+ssl closer-mop (sbcl version checked) ironclad (https://github.com/sharplispers/ironclad/compare/v0.54...v0.55) iterate (https://gitlab.common-lisp.net/iterate/iterate/-/commit/d5f83a9985d7f22c786a95016ba2aa275195baa2) postmodern (https://github.com/marijnh/Postmodern/compare/v1.32.8...v1.32.9) quri (https://github.com/fukamachi/quri/commits/master) salza2 trivial-features (https://github.com/trivial-features/trivial-features/compare/v0.9...v1.0) uax-15 (https://github.com/sabracrolleton/uax-15/commits/master)
Maintainer informed via mail
Hi Mario, Thanks for the patch. Please do not add '# $FreeBSD$' at the beginning of the Makefile anymore. This is obsolete after switching the ports tree from svn to git. BTW, have you already converted your ports tree? Your patch also accidentally contains a 'Makefile~' file ;) Setting the PORTREVISION is to force the rebuild of even already installed packages to use the 'trivial-features upgrades in endianess', right?
(In reply to Rainer Hurling from comment #2) Hi Rainer, thanks for your quick reply! Yes, the VC head doesn't make much sense anymore with git, thanks for pointing that out! I will put a bit more time into taking care of the provided patch so no backup files will be included (usually I have Makefile with a "phony clean" target). Speaking of time: I have built some maintenance tools to maintain quicklisp ports like this one. I've started from scratch again as someone asked me to write about "why I love POSIX/SUS" and I wrote it in that manner only. But it's not fully automated and so I decided to have some quarterly refresh of that port. I've gone through all diffs and looked into the release notes if available and I decided that it would be a good idea to increase the PORTREVISION. It might be interesting to have a flag that marks this port as "update ports Makefile but leave binary"; okay, I could just have let PORTREVISION untouched (i.e. 0) but the changes were a bit more than just cosmetic and I couldn't tell if they would deeply affect parts I wouldn't think of (this port needs almost 70 distfiles) ). So what do you think: * should we proceed and increase the PORTREVISION as proposed * or should we proceed and remove the PORTREVISION from the new Makefile to offer a new port but not to build it automatically * or leave v3.6.2 as is to wait for some major updates? I will include the patch just after this reply for the case they'll be needed without the backup file, sorry for that and thank you very much again for your comments!
Created attachment 224822 [details] v3.6.2-1 see comment
Hi Mario, your updated patch contains that '# $FreeBSD$' line again ;) You should not diff against a svn version ... Thank you for the detailed presentation about the possible procedures with the PORTREVISION etc. If it were my port, I would probably just bump the PORTREVISION as already present in the patch. My first test on Poudriere with your new patch breaks in the build with the following output: [00:04:19] [01] [00:00:35] Status databases/pgloader3 | pgloader3-3.6.2_1: build Type HELP for debugger help, or (SB-EXT:EXIT) to exit from SBCL. restarts (invokable by number or by possibly-abbreviated name): 0: [CONTINUE] Skip this save hook. 1: [RETRY ] Retry EVAL of current toplevel form. 2: Ignore error and continue loading file "/wrkdirs/usr/ports/databases/pgloader3/work/pgloader-3.6.2/dumper-2SKVI5f7.lisp". 3: [ABORT ] Abort loading file "/wrkdirs/usr/ports/databases/pgloader3/work/pgloader-3.6.2/dumper-2SKVI5f7.lisp". 4: Ignore runtime option --load "/wrkdirs/usr/ports/databases/pgloader3/work/pgloader-3.6.2/dumper-2SKVI5f7.lisp". 5: Skip rest of --eval and --load options. 6: Skip to toplevel READ/EVAL/PRINT loop. 7: [EXIT ] Exit SBCL (calling #'EXIT, killing the process). (SB-INT:CALL-HOOKS "save" (#<FUNCTION CLOSE-FOREIGN-LIBS>) :ON-ERROR :ERROR) 0]
I found another hint in the Poudriere log file. Obviously there is a problem with libssl: [..snip..] While evaluating the form starting at line 29, column 0 of #P"/wrkdirs/usr/ports/databases/pgloader3/work/pgloader-3.6.2/dumper-2SKVI5f7.lisp": debugger invoked on a SIMPLE-ERROR in thread #<THREAD "main thread" RUNNING {10015481B3}>: Problem running save hook #<FUNCTION CLOSE-FOREIGN-LIBS>: dlerror() returned an error while trying to close shared object "libssl.so": "Undefined symbol \"SSL_CTX_set_tmp_rsa_callback\"" debugger invoked on a SIMPLE-ERROR in thread #<THREAD "main thread" RUNNING {10015481B3}>: Maximum error nesting depth exceeded * CORRUPTION WARNING in SBCL pid 32822 pthread 0x800a12000: Memory fault at 0x8021501c0 (pc=0x8021501c0, fp=0x8011ff240, sp=0x8011ff238) pthread 0x800a12000 The integrity of this image is possibly compromised. Continuing with fingers crossed. # that's ugly, but necessary when building on Windows :( mv build/bin/pgloader.tmp build/bin/pgloader mv: rename build/bin/pgloader.tmp to build/bin/pgloader: No such file or directory gmake[1]: *** [Makefile:149: build/bin/pgloader] Error 1 gmake[1]: Leaving directory '/wrkdirs/usr/ports/databases/pgloader3/work/pgloader-3.6.2' *** Error code 1 Stop. make: stopped in /usr/ports/databases/pgloader3 =>> Cleaning up wrkdir ===> Cleaning for pgloader3-3.6.2_1 build of databases/pgloader3 | pgloader3-3.6.2_1 ended at Tue May 11 07:20:51 CEST 2021 build time: 00:15:55 !!! build failure encountered !!!
(In reply to Rainer Hurling from comment #6) (In reply to Rainer Hurling from comment #5) It's so embarrassing, Rainer, I'm going to meditate on it. In the lotus seat. Sitting on my obviously snuffed-out build host. I'll close that and put you CC when I'm ready, if that's okay. Thank you for your helpful comments. :)
s. comments