summaryrefslogtreecommitdiffstats
Commit message (Collapse)AuthorAgeFilesLines
* 8.1238.123Thierry Vignaud2020-05-013-3/+3
|
* fix running genhlist2 with the right perl versionThierry Vignaud2020-05-012-0/+4
| | | | | | | | | | | | | | | | | | | | Else genhdlist2 may fail to load URPM if it doesn't use the perl version for which URPM was installed. This affects CPAN smokers where several perl versions are installed. Eg in http://www.cpantesters.org/cpan/report/6808b728-8a34-11ea-bab2-e37d1f24ea8f CPAN says INC= /usr/perl5.30.0p/lib/site_perl/5.30.0/amd64-freebsd /usr/perl5.30.0p/lib/site_perl/5.30.0 /usr/perl5.30.0p/lib/5.30.0/amd64-freebsd /usr/perl5.30.0p/lib/5.30.0 Which is consistent with $^X = /usr/perl5.30.0p/bin/perl5.30.0 But genhdlist2 fails with: Can't locate URPM.pm in @INC (you may need to install the URPM module) (@INC contains: /usr/home/cpansand/.cpan/build/2020042914/urpmi-v8.122-1/blib/lib /usr/home/cpansand/.cpan/build/2020042914/urpmi-v8.122-1/blib/arch /usr/perl5.28.2-RC1p/lib/site_perl/5.28.2/amd64-freebsd /usr/perl5.28.2-RC1p/lib/site_perl/5.28.2 /usr/perl5.28.2-RC1p/lib/5.28.2/amd64-freebsd /usr/perl5.28.2-RC1p/lib/5.28.2 .) at /usr/home/cpansand/.cpan/build/2020042914/urpmi-v8.122-1/t/genhdlist2 line 6. BEGIN failed--compilation aborted at /usr/home/cpansand/.cpan/build/2020042914/urpmi-v8.122-1/t/genhdlist2 line 6. Ie it used the wrong perl-5.28.2 instead of perl-5.30.0
* replace "xxx\"xxx" with qq(xxx"xxx): no need to escape <">Thierry Vignaud2020-05-011-1/+1
|
* BSD: prevent failing due to broken deps generatorThierry Vignaud2020-05-012-0/+4
| | | | | | | | | | | | | some FreeBSD CPAN smokers sometimes fails with: error: Couldn't exec /usr/local/lib/rpm/elfdeps: No such file or directory The testsuite passes smoothly before & after that change. The only impact is that with that, t/media/rpm-query-in-scriptlet/rpm-query-in-scriptlet-1-1.x86_64.rpm lose one requires on one of its own provides: removed REQUIRES librpm.so.9()(64bit) Which is fine.
* factorize a testThierry Vignaud2020-05-011-2/+4
| | | | It's more readable and will help next commit
* just use only one variableThierry Vignaud2020-05-011-4/+4
| | | | Thus we got a shorter command which will help for next commits
* perl_checker cleanupThierry Vignaud2020-05-011-1/+1
|
* bump requires on URPM to 5.28Thierry Vignaud2020-05-012-1/+2
| | | | so that we can be tester nearly everywhere
* fix testuite if glibc-static-devel isn't installedThierry Vignaud2020-05-012-0/+3
| | | | On some older distributions (eg: Mageia 2), it would fail.
* fix running testsuite with rpm-4.9Thierry Vignaud2020-05-012-1/+7
| | | | | Basically restoring commit 654ffaa9caf78556b9759c18345b5ec9207f3c43 for older rpm
* do not skip network tests on pre-UsrMove distrosThierry Vignaud2020-05-012-1/+4
| | | | | Eg: on Mageia 2, the network test would be skipped due to faulty test which caused us to think no downloader would be available
* 8.1228.122Thierry Vignaud2020-04-294-3/+5
|
* cpan_testers: fallback to bundled genhdlist2Thierry Vignaud2020-04-297-0/+1118
| | | | | we could have done it for all tests directly from within t/helper.pm, but only those tests actually need to find genhdlist2
* do not restrict PATH during testsuiteThierry Vignaud2020-04-292-1/+3
| | | | (needed for next commit)
* retry if curl failed with timeoutThierry Vignaud2020-04-292-1/+17
|
* explain the relationship with rpmtoolsThierry Vignaud2020-04-291-0/+4
|
* explain the relationship with URPMThierry Vignaud2020-04-292-0/+3
|
* prevent forking a useless shellThierry Vignaud2020-04-291-1/+4
| | | | as shown by previous commit
* silent rpm --version during installation/removalThierry Vignaud2020-04-292-2/+6
| | | | | | | | | | | | | | | | Calling "LC_ALL=C rpm --version" from urpm::install::install_logger() sometimes causes the following to be displayed when using --urpmi-root: "shell-init: error retrieving current directory: getcwd: cannot access parent directories: No such file or directory" Typically, we run into: chdir("/") = 0 chroot("/home/tv/mga/pkgs/RPM/perl-URPM/mga7/") = 0 (...) clone() execve("/bin/sh", ["sh", "-c", "LC_ALL=C rpm --version"], ... getcwd("(unreachable)/", 4095) = 15 So just caches the rpm version prior to the first transaction and voila
* fix failing when serving over HTTP from within /rootThierry Vignaud2020-04-292-2/+7
| | | | | | | | | rationale: If we switch from root to nobody/nogroup while running in /root, everytime simple-httpd accesses to a files while answering to urpmi will fail with "-1 EACCES (Permission denied)" Then t/tmp/error.log contains "Error - 404 - file not found" but that's hidden when running tests as part of CPAN testers.
* %find_lang sometimes fails on FreeBSDThierry Vignaud2020-04-294-0/+19
| | | | | + /usr/local/lib/rpm/find-lang.sh /usr/home/cpansand/.cpan/build/2020042613/urpmi-v8.121.20-6/t/tmp/BUILDROOT/various-1-1.x86_64 test-various error: Bad exit status from /usr/home/cpansand/.cpan/build/2020042613/urpmi-v8.121.20-6/t/tmp/rpm-tmp.kKrrdT (%install)
* display RPM version in logsThierry Vignaud2020-04-292-0/+3
|
* skip rpm-query-in-scriptlet.spec on FreeBSDThierry Vignaud2020-04-292-1/+13
| | | | | | | | | | | | | | | | | | | because of: + cd / + cpio -pumdL /usr/home/cpansand/.cpan/build/2020042509/urpmi-v8.121.15-2/t/tmp/BUILDROOT/rpm-query-in-scriptlet-1-1.x86_64 cpio: Can't stat /bin/sh: cpio: Can't stat /usr/local/bin/rpm: 24035 blocks error: Bad exit status from /usr/home/cpansand/.cpan/build/2020042509/urpmi-v8.121.15-2/t/tmp/rpm-tmp.Q4A94o (%install) See related debuging in: http://www.cpantesters.org/cpan/report/e7284484-878b-11ea-a99d-f0811f24ea8f + grep / list + ls -l /usr/local/lib/libnss3.so /bin/sh /bin/sh: /lib/libedit.so.7 /lib/libc.so.7 /lib/libncursesw.so.9 /usr/local/bin/rpm /usr/local/bin/rpm: /usr/local/lib/librpm.so.9 /usr/local/lib/librpmio.so.9 /usr/local/lib/libdb-5.3.so.0 /usr/local/lib/libnss3.so /usr/lib/libbz2.so.4 /lib/libz.so.6 /usr/local/lib/libpopt.so.0 /usr/lib/liblzma.so.5 /usr/local/lib/libzstd.so.1 /usr/local/lib/liblua-5.2.so /lib/libm.so.5 /lib/libthr.so.3 /usr/local/lib/libintl.so.8 /lib/libc.so.7 /usr/local/lib/libnssutil3.so /usr/local/lib/libplc4.so /usr/local/lib/libplds4.so /usr/local/lib/libnspr4.so /usr/lib/libdl.so.1 ls: /bin/sh:: No such file or directory ls: /usr/local/bin/rpm:: No such file or directory error: Bad exit status from /usr/home/cpansand/.cpan/build/2020042520/urpmi-v8.121.17-2/t/tmp/rpm-tmp.GcQOIE (%install)
* split some variables for readabilityThierry Vignaud2020-04-291-1/+3
|
* cpan-testers: fix failing to build some rpms on FreeBSDThierry Vignaud2020-04-292-1/+4
| | | | | | | | | | | | | | | | | | | | | unset %__os_install_post fixes to fix FreeBSD failures (RT#131162) %__os_install_post cause build failures on FreeBSD due to rpm not being adapted: such as http://www.cpantesters.org/cpan/report/0a319070-86d8-11ea-a1f7-ec411f24ea8f: + /usr/local/lib/rpm/brp-compress /usr/local + /usr/local/lib/rpm/brp-strip /usr/bin/strip + /usr/local/lib/rpm/brp-strip-static-archive /usr/bin/strip xargs: illegal option -- d usage: xargs [-0opt] [-E eofstr] [-I replstr [-R replacements] [-S replsize]] [-J replstr] [-L number] [-n number [-x]] [-P maxprocs] [-s size] [utility [argument ...]] error: Bad exit status from /usr/home/cpansand/.cpan/build/2020042509/urpmi-v8.121.15-0/t/tmp/rpm-tmp.P9J8JW (%install) Stripping also often fails on FreeBSD: /usr/bin/strip: unable to copy file '...BUILDROOT/arch_to_noarch-1-1.x86_64/usr/lib/test-arch_to_noarch/ldconfig' reason: Permission denied As a side effect, it also fixes: Error: Could not execute /usr/local/lib/rpm/elfdeps: file or directory not found
* ignore .gitThierry Vignaud2020-04-291-2/+2
|
* use a more neutral C compiledThierry Vignaud2020-04-292-2/+4
| | | | | | Aka adapt to a world where gcc is no more the apex predator... Several Linux distros as well as FreeBSD >= 10 default to clang instead of gcc.
* simplify: BSD support -L but not --dereferenceThierry Vignaud2020-04-292-12/+2
| | | | | | | This might help the following failure: http://www.cpantesters.org/cpan/report/c8c690aa-8587-11ea-81c4-93681f24ea8f cpio: Can't lstat /bin/sh: cpio: Can't lstat /usr/local/bin/rpm:
* do not fork (daemonize is implied by {setsid})Thierry Vignaud2020-04-292-1/+1
| | | | | | | Thus fixing killing simple-httpd when test completes, which was broken since basically forever. This prevents dummy testsuite faillures when performing parallel runs on different perl/rpm versions (eg: using chroots).
* fix "The plan was already output"Thierry Vignaud2020-04-291-4/+6
| | | | | | | | | | | The error message being in some conditions: "The plan was already output at /usr/lib/perl5/5.22.3/Test/Builder.pm line 2510." We need both else we got: "Tests were run but no plan was declared and done_testing() was not seen." also fix inverted condition :-(
* fix a warning causing Test::More to think it failed when /proc is not mountedThierry Vignaud2020-04-292-2/+9
| | | | | | | | | | | | | | | | | | | You tried to plan twice at t/superuser--should-restart.t line 14. t/superuser--should-restart.t ................................... skipped: (no reason given) (...) Test Summary Report ------------------- t/superuser--should-restart.t (Wstat: 65280 Tests: 0 Failed: 0) Non-zero exit status: 255 Files=40, Tests=3901, 801 wallclock secs ( 0.96 usr 0.09 sys + 174.18 cusr 30.16 csys = 205.39 CPU) Result: FAIL Failed 1/40 test programs. 0/3901 subtests failed.A On the other side we want to make it work in both cases so we end doing that.
* check for genhdlist2 --help tooThierry Vignaud2020-04-291-0/+4
|
* cpan_testers: show verbosely why genhdlist2 do not workThierry Vignaud2020-04-292-0/+8
|
* add more extensive testsThierry Vignaud2020-04-292-0/+6
|
* cpan-testers: work without glibc-static-develThierry Vignaud2020-04-292-3/+18
| | | | | | just build a dynamic binary but exclude its deps rationale: quite a lot of CPAN testers do not have static libs installed...
* explain why we need our own "static shell"Thierry Vignaud2020-04-291-0/+2
| | | | historically we were used static dash but that was too Mageia centered
* describe the testsuiteThierry Vignaud2020-04-292-0/+30
|
* disable fsync/fdatasync while running testsThierry Vignaud2020-04-292-0/+3
|
* cpan-testers: fix failing test on Debian 9/FreeBSDThierry Vignaud2020-04-282-2/+3
| | | | | | commit 1089a6c113c53cd77147808ebeb4f6d0ef03fa89 really was about silencing a warning, so let's not break other OSes who do not have that lib or not in this location
* Also look at $Config{bin} for genhdlist2Thierry Vignaud2020-04-281-1/+2
| | | | | | | | | | It should help http://www.cpantesters.org/cpan/report/9e067a66-83f1-11ea-85a9-af7f1f24ea8f which fails. That smoker had rpmtools/URPM already installed normally into this perl's sitedir & bindir. Adding the blib/script directories didn't help that case. So try $Config{bin} too
* try harder to find genhdlist2 on CPAN testersThierry Vignaud2020-04-282-0/+7
| | | | | | | | | | | | rationale: in some cases, all dependency modules and scripts still reside in the uninstalled blib directories. CPAN.pm knows how to deal with uninstalled modules (all the blib/lib directories go into the PERL5LIB environment variable). But it does not do something similar for the blib/script directories. So we have to try harder to find genhdlist2 by adding those blib/script directories to PATH if they exist
* split set_path()Thierry Vignaud2020-04-282-3/+7
|
* fix test with rpm < 4.13 on CentOS7Thierry Vignaud2020-04-282-1/+2
|
* perform some file conflict tests only on MageiaThierry Vignaud2020-04-282-12/+31
| | | | | rationale: Mageia's rpm is patched in order to not conflict for doc files:
* introduce is_mageia()Thierry Vignaud2020-04-282-2/+6
|
* check & report where genhdlist2 isThierry Vignaud2020-04-282-0/+5
|
* describe more tests to help pinpoint regressionsThierry Vignaud2020-04-2810-25/+26
|
* make test output clearerThierry Vignaud2020-04-281-1/+3
|
* check that web server has actually startedThierry Vignaud2020-04-283-0/+13
| | | | reports its error messages if any
* perl_checker cleanupsThierry Vignaud2020-04-281-1/+1
|