[Pkg-rust-maintainers] Bug#1071597: rust-laurel - autopkgtest failure on s390x

[Peter Green]

Package: rust-laurel
Version: 0.6.2-1
Severity: serious

rust-laurel's autopkgtest fails on s390x. I belive the patch
skip-parse_syslog-on-big-endian.patch should be reinstated
but I do not want to get into a revert war with the
maintainer.

So I feel I need to lay out, in more detail than
is visible in the changelog and git history, what I have
done regarding this package and why.

My main role in the rust team has been trying to keep the
"greater whole" of rust packages in a consistent state and
moving forward. Since upstream rust dependencies tend to be
pessimistic and Debian frowns upon packaging multiple
versions of the same software, this inevitablly involves
patching a bunch of packages.

I use autopkgtests as a tool to help minimize the chances
that these changes cause undetected breakage. As such when
bumping a dependency in a package that has no functional
autopkgtests, I will usually try to get at least some test
coverage.

Regarding rust-laurel specifically, up to version 0.5.6-1,
the tests were skipped.

In version 0.5.6-2, as part of preparing for the nix 0.27
update, I patched rust-laurel to get tests running. The
tests passed in my local tests on amd64 and so I uploaded
it.

Tests on debci revealed some failures though, there was
a test failure on 32-bit systems due to an integer
overflow in some time calculation code, and a failure
on s390x architectures which I determined to be down
to endian-specific test date in the test.

I prepared a fix for the time calculation code, which I
also posted upstreamed.

I decided that fixing the test data would probablly not
have a positive effort/benefit ratio and therefore added
a patch to skip the test on big-endian architectures. This
was still an improvement over the prior situation
where no tests were being run at all.

I uploaded these changes as 0.5.6-2.

The time overflow issue was fixed upstream in versions
0.6.0 and later. When Hilko uploaded version 0.6.1-1
he dropped my patches. The time overflow issue was fixed
upstream, so this made sense but nothing had been done
upstream to address the big endian issue. So this lead
to the tests once again failing on s390x.

I figured that this was a mistake, that Hilko had
incorrectly assumed that the big endian issue had been
taken care of upstream and after nearly a month of the
package sitting unable to migrate to testing, I
reinstated the patch and tweaked it to apply to the
new upstream version.

I made further uploads of 0.6.1 to address issues
related to the time64 and nix transitions.

When preparing the upload of 0.6.2-1, Hilko once again
dropped the patch with a changelog entry of
"Drop unneeded patches".

So that brings us to where we are now, the package
fails it's autopkgtests on s390x and I do not feel
it's appropriate to reinstate the patch for the second
time without first trying to get some feedback from
the maintainer.