On Wed, 19 Mar 2025 at 18:10:34 -0100, Graham Inggs wrote:
It seems the BTS considers that 4.3.0-2 will re-introduce this bug to testing because the changelog entry for 4.2.1-4 above was dropped from debian/changelog.
As far as I can see, the BTS is correct: there are changes from 4.2.1-4 and 4.2.1-5 listed in <https://salsa.debian.org/science-team/mpich/-/commits/debian%2Ftrixie?ref_type=heads> that are not present in <https://salsa.debian.org/science-team/mpich/-/commits/debian%2Flatest?ref_type=heads>, and were therefore lost when 4.3.0 was re-uploaded from experimental into unstable shortly before the toolchain freeze. Can debian/trixie be merged into debian/latest to reinstate those changes? (It also seems misleading that there is a debian/trixie branch but it does not correspond to either the version in trixie, or the version that is waiting to migrate into trixie.) It looks to me as though #1088919 may have been essentially the same issue as #1102928, but in the other branch. Fixing at least mpi-c.pc would likely allow valgrind to be built successfully on 32-bit (#1103970) which in turn would allow the unrelated RC bug fix #1057621 to migrate to testing. smcv
