> the fix for PR ada/117996 causes a change for the gnat-13-<hash> > provides, causing uninstallability for all ada packages. This fix > doesn't look like an ABI change, neverless it triggers one on the > packaging side. > according to the tracker, https://tracker.debian.org/pkg/gcc-13 > there are no abi changes on ppc64el. why?
No idea. I would also not expect these changes to break the ABI. > If I understand it correctly, we still plan to switch to gnat 14 for > the upcoming trixie release, but this will be a problem for > backports, and also assuming we had the switch to gnat 14 already > done, we would have the problem there again. Is that hash too > strict, or is this a real issue? The language requires that all reverse dependencies are rebuilt after any change in the Ada sources of a library, so the hash is correct. Until gnat-12, a change in libgnat was unnoticed until someone attempts to rebuild an Ada package, and the fix was to rename all Ada -dev packages. With the hash/Provides mechanism, such a change is immediately visible as autopkgtest failures, and can be fixed with a bin-NMU of all Ada libraries in proper order (this remains an option). However, I see that you have reverted the fix for PR ada/117996 from libgnat-13, and it make sense to only apply it to gcc-13 after the gnat-14 transition.. It will then affect Ada libraries neither in bookworm backports (gnat-12) nor in trixie (gnat-14). >From its description, I suspect that PR ada/117996 does not affect many users in testing/unstable.
