On Thu, Mar 19, 2026 at 01:33:27PM +0100, Michael Banck wrote: > On Thu, Mar 19, 2026 at 01:27:20PM +0100, Santiago Vila wrote: > > If it's unsupported in trixie, why do we still include it in trixie? > > I don't think anybody said it was unsupported. The FTBFS you filed > seems to show some drift/flakeyness in the test suite while the code > actually builds fine.
Ok, for completeness: This worked ok in Debian 13.2 and it started to fail in Debian 13.3. Based on the dates, this seems to be triggered by the upgrade in trixie from postgresql 17.6-0+deb13u1 to 17.7-0+deb13u1). Also, to be sure, I've now built the package 100 times and it failed 100 times, so it's not flaky. It seems to always fail in the same way. So I guess it falls under the category of drift (as in "the real output does not match anymore the expected output"), but we still don't know why. > If you find a real bug in the way pgl-ddl-deploy works, I think we are > happy to take a look at it. Well, is not that precisely what build-time tests are for? They indicate potential bugs in the code, but until somebody looks at them, we will never know if it's a bug in the code or a bug in the tests. In my experience, a package can be completely broken in trixie for a long time without anybody noticing. For example, bug #1128265 in policyd-rate-limit comes to mind. In this case the package had no tests at all, which surely helped the bug to remain unnoticed. Thanks.
