Am / On Tue, 9 Mar 2021 16:35:04 -0300 schrieb / wrote Antonio Terceiro <terceiro@debian.org>: > > Now every time I push changes to that > > mentors/webgen-git I seem to trigger some build pocess > > of the package and the funny thing is, that this build > > process has a totally different outcome than the one I > > initiate on my local git with the meta-build script: > > > > When I run the build-script on my local machine the > > build in the (unstable) sbuild environment works fine > > (including all tests during the build process) and all > > tests in the lxc-unstable environment run fine as well. > > > > The build-process that is initiated on salsa fails > > because some tests (that are done during the > > build-process) fail. And these are the same tests that > > run fine here. > > > > Can anybody explain me this difference? > > Without any details, there is no way for someone to help > you. This is equivalent to a bug report that just says > "foo doesn't work!". > > Where are the logs that show the failure? Can you > identify a specific error message? You're right of course! I thought there was a general difference in the build processes which would explain the difference. And I think there is. Looking into the logs of both builds I see that my local build-process does the tests with GEM_PATH= ... :/usr/lib/x86_64-linux-gnu/ruby/gems/2.7.0: ... /usr/lib/x86_64-linux-gnu/rubygems-integration/2.7.0 -S rake -f debian/ruby-tests.rake and with: Run options: -v --seed 41315 The salsa build process does the tests with: GEM_PATH= ... :/usr/lib/ruby/gems/2.7.0:/usr/lib/i386-linux-gnu/ruby/gems/2.7.0: ... /usr/lib/x86_64-linux-gnu/rubygems-integration/2.7.0 -S rake -f debian/ruby-tests.rake and with: Run options: -v --seed 63101 So the salsa tests are running on an i386-system, mine on an amd64. In the result the tests on my machine are running fine, only one is skipped: "212 runs, 2102 assertions, 0 failures, 0 errors, 1 skips" and the package is built; On salsa I get: "212 runs, 1501 assertions, 16 failures, 29 errors, 1 skips rake aborted!" and also dh_auto_install aborts. But I thought ruby packages were architecture independent? The log of the salsa build and the list the 16 failures and 29 errors are here: https://salsa.debian.org/ruby-team/mentors/webgen/-/jobs/1495406/raw The first is an Argument error in TestErb#test_static_call: "invalid byte sequence in US-ASCII" followed by many with: "Can't have two nodes with same alcn: /file.html" Because in webgen's alcn often only differ by a language extension (e.g.: file.de.html vs. file.en.html), the first test-failure might be the reason for all of them. But to be honest, that's more feeling than knowing. So, any idea from you? Or do I have to contact the program's author once more? Klaumi ----- Klaumi Klingsporn mail: klaumikli@gmx.de web: www.klaumikli.de
Attachment:
pgppuSw7QCQsN.pgp
Description: Digitale Signatur von OpenPGP