Dear maintainers,
Following up on my earlier message about taking over approx (bug #948603), I
have set up a Debian sid build environment and reproduced the current
packaging. I am writing to report what I found, because two of my three initial
concerns turned out to be non-issues and I would rather correct the record than
leave misleading impressions.
Summary of findings:
1. pristine-tar: not broken. My initial report that pristine-tar was broken was
my own mistake. I had run "pristine-tar gentar" from a directory outside the
git repository; pristine-tar needs the git tree to rebuild the delta base and
fails there. Run from inside the repository it works perfectly and reproduces
approx_5.13.orig.tar.gz byte-for-byte identically to the copy in the Debian
archive. No issue here.
2. systemd-tmpfiles dependency: this one is real. approx ships
/usr/lib/tmpfiles.d/approx.conf via debian/approx.tmpfiles. Rebuilding approx
from its current git master against sid produces a binary whose Depends field
lacks "systemd | systemd-standalone-tmpfiles | systemd-tmpfiles", whereas
approx 5.13-3+b1 currently in sid has it. The cause is in debhelper, not in
approx: dh_installsysusers adds the sysusers alternative through
${misc:Depends}, but dh_installtmpfiles contains no addsubstvar() call, even
though its manpage states that it adds the dependency. I have prepared the
one-line debian/control fix and verified that the resulting Depends field is
then identical to the official binary. I have not pushed anything.
3. Source format: no change needed. debian/source/format is 3.0 (quilt) while
debian/gbp.conf sets pristine-tar = True. I initially read this as an
inconsistency, but it is in fact the standard combination: the OCaml team uses
git-buildpackage with pristine-tar for upstream tarballs, and pristine-tar
exists precisely to supply the .orig tarball that 3.0 (quilt) requires, with
bit-identical checksums for dak. I am leaving it alone.
Two questions:
- Should I file a separate bug against debhelper about the dh_installsysusers /
dh_installtmpfiles asymmetry? I have a draft ready including reproduction
steps, and I would rather not duplicate a report someone may have already
filed. I would also appreciate knowing whether the maintainers prefer that
affected packages declare the dependency explicitly in the meantime.
- Lintian reports "missing-systemd-timer-for-cron-script" for
/etc/cron.weekly/approx. This is pre-existing, as the official binary in sid
reports the same warning, so it is not a regression. Would the team prefer it
to be addressed now, by converting the cron script into a systemd timer, or
left as-is? I can prepare a Merge Request either way.
I have a fix/blockers-wip branch prepared locally with the debhelper workaround
and an UNRELEASED changelog entry, but I have deliberately not pushed anything
and will hold off on a Merge Request until I hear from you about maintainership.
Happy to provide the debdiff or the draft debhelper bug report if useful.
Best regards,
Leonardo
