On Sun, May 31, 2026 at 06:03:13PM -0500, Steven Robbins wrote: > Also - sometimes the tests.reproducible-builds.org status reports FTBFS > while > reproduce.debian.net has a checkmark and the regular debian build machines > appear OK. What's different in the reproducible-builds setup?
Hi. I'd like to point out that you don't necessarily need to have two different build setups to have a build failure in one of them and not the other. For example, if your package has a flaky test which makes the build to fail 50% of the time, this will happen easily. This will certainly be an incentive for maintainers to fix or disable them before they make more harm. While we are at it: The aim of reproducible-builds is to allow anybody to verify that the .deb may be reconstructed identical to the one in the archive. But this will not happen if the package only builds 50% of the time (see above), or it only builds in certain cases that we can't express via build-dependencies in debian/control. Therefore, in line with requiring packages to be reproducible, I would also expect Release Managers to be more strict about build failures which do not happen in buildd.debian.org. We want everybody to be able to build packages, not just buildd.debian.org. Thanks.
