Kai Pastor, DG0YT: > Am 13.11.2017 um 20:15 schrieb Adrian Bunk: >>> I didn't find a recommendation how to solve the issue. I hope a custom macro >>> is okay. >> Based on the discussion in #876901 [1] it is still unclear how this >> should be resolved in general. >> > One more remark: > > Replacing __FILE__ with an individual macro makes it much harder to detect > embedded build paths in source code review. Proliferation of individual > macros could become worse than __FILE__ IMO. > > > I see the discussion on debian-qt-kde. The proposal in > https://lists.debian.org/debian-qt-kde/2017/11/msg00110.html: > >> If all the test files reside underneath the same directory, like tests/, you >> could: > >> 4. export >> BUILD_PATH_PREFIX_MAP="$BUILD_PATH_PREFIX_MAP=tests=$BASEDIR/tests". > > will probably worked for this package (old versions). But such information > needs to be in documentation, at the time such a change is introduced. In > addition, there needs to be much better correlation of changes in the > environment and changes in reproducibility build success. It seems there are > two variations tested, but __FILE__ vs. not __FILE__ is not included. > > I remember that I spent time studying changelogs of gcc and Qt packages in > Debian when this bug was opened, but I was unaware of the particularities of > the gcc in reproducible builds. The rbuild log that I find online does not > contain the exact c++ compiler package name and version. What an irony, the > reproducible-builds logs do not provide enough information to consider them > reproducible. Even Debian package build logs on build.opensuse.org offer more > details. We could have solved this much earlier, and without wasting CPU > cycles and developer time. >
There was some mis-co-ordination, Adrian did not tell us he was filing the bug, we the reproducible-builds team were not intending to file a bug, and I missed the first few messages of the bug report so I assumed he already explained that the FTBFS is only a problem on tests.r-b.org and not the official Debian archive or buildds. The patch currently being used on tests.r-b.org is here: https://anonscm.debian.org/cgit/reproducible/debrepatch.git/tree/toolchain-patches/gcc-7_BPPM.patch The documentation for the variable seems to have been chopped off by mistake, but it is in the patch that I sent upstream to GCC, ctrl-F for "invoke.texi": https://gcc.gnu.org/ml/gcc-patches/2017-07/msg01318.html Indeed, the exact version of the C++ compiler does not appear in the rbuild: https://tests.reproducible-builds.org/debian/rb-pkg/unstable/amd64/openorienteering-mapper.html That is because schroots are installed with "build-essential" by default and so gcc doesn't get downloaded every single time for each package build. That's an unfortunate situation and I can see it's ironic, you could probably file a bug against jenkins.debian.org or perhaps pbuilder to have the builders dump this information into the log, even if the build fails. X -- GPG: ed25519/56034877E1F87C35 GPG: rsa4096/1318EFAC5FBBDBCE https://github.com/infinity0/pubkeys.git
