Hi, On 2024-12-22 10:30, Lucas Nussbaum wrote: > During a rebuild of all packages in sid, your package failed to build > on amd64.
[...] > > 115 tests, 1 failures. During an archive rebuild with both GCC 16 and GCC 15 on arm64, I found that the tests (very) occasionally pass. Here the logs for a build failure: https://people.debian.org/~ema/gcc-16-rebuilds/output-1/open-adventure_arm64.build.xz And success: https://people.debian.org/~ema/gcc-16-rebuilds/output-2/open-adventure_arm64.build.xz This is actually bad news, given that flaky tests increase the amount of work needed when doing comparative archive rebuilds. The procedure is as follows, for instance when looking for regressions moving from GCC 15 to GCC 16: - Build the archive with GCC 16 - Rebuild failing packages GCC 15 - File bugs for packages building fine with 15 but failing with 16 Packages such as open-adventure may fail building with GCC 16 but succeed with GCC 15, even though the problem is not due to the compiler at all. A reliable ftbfs would be much better! I'll start usertagging bugs in this category as debian-qa@ / flaky-ftbfs to highlight the distinction from the nice ftbfs we can rely on. :-) ema
