[Debian-med-packaging] Bug#1148077: src:diamond-aligner: fails to migrate to testing for too long: FTBFS on s390x

[Sebastien Bacher]

Control: tags -1 + patch fixed-upstream

Hi,

The s390x FTBFS in 2.2.6-1 is a big-endian bug, and upstream has
already fixed it:


https://github.com/bbuchfink/diamond/commit/093dafab9f9ecf2a0666ddd0dda48e9eff29b7c6
  (https://github.com/bbuchfink/diamond/pull/986, merged 2026-09-26, not
  in any release yet)

DatabaseFile::is_diamond_db() reads the .dmnd magic number as raw bytes
and does not convert it from the on-disk little-endian format. On s390x
the comparison fails, so diamond treats the database as FASTA, and the
build-time test fails with:

  Opening the database... Error: Error opening file test/C.faa.diamond. No
such file or directory

The commit adds the missing byteswap there and in read_seqinfo().

We have applied it in Ubuntu on top of 2.2.5-2, together with a backport
of the 2.2.6 generic FingerPrint code. The s390x build now passes, and
all 20 diamond blastp runs in debian/tests/diamond-calls complete.

The commit applies to 2.2.6 unchanged; it is the same code that Ubuntu
tested on 2.2.5. Adding it as a patch to 2.2.6-1 should fix s390x and
unblock the migration.

The patch with a DEP-3 header is attached.

Cheers,
Sebastien
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://alioth-lists.debian.net/pipermail/debian-med-packaging/attachments/20261008/cc8d489b/attachment-0001.htm>
-------------- next part --------------
A non-text attachment was scrubbed...
Name: big-endian-dmnd-byteswap.patch
Type: text/x-patch
Size: 1698 bytes
Desc: not available
URL: <http://alioth-lists.debian.net/pipermail/debian-med-packaging/attachments/20261008/cc8d489b/attachment-0001.bin>