Re: Potential MBF: packages failing to build twice in a row
On Sat, 05 Aug 2023 at 17:06:27 +0200, Lucas Nussbaum wrote:
> Should we give up on requiring a 'clean' target that works? After all,
> when 17% of packages are failing, it means that many maintainers don't
> depend on it in their workflow.
I think it's somewhat inevitable that code paths that aren't frequently
exercised don't work. If a majority of maintainers are doing all of
their builds with git-buildpackage, or dgit --clean=git, or something
basically equivalent to one of those, then `debian/rules clean` will
never actually be run against a built tree. For teams with a strongly
preferred workflow (like the Perl, Python and GNOME teams consistently
using git-buildpackage), this seems particularly likely.
I think we need to think about what benefit this Policy requirement brings
us, and whether it's worth the cost, before treating it as important:
the higher we choose to make the cost of fixing this class of bug,
the higher the bar should be for treating it as a real bug at all,
or treating it as RC.
For me, the main purpose of `debian/rules clean` is being able to do
incremental builds while debugging something - but if I want to do
incremental builds, it's quite likely that I'll also be using
`debuild -b -nc` to make the builds genuinely incremental (and then a fully
clean build from first principles at the end, to verify that whatever issue
I'm debugging is really fixed).
Having looked at some of the packages with my name on in
all_failing.txt.dd-list, many of them are a simple matter of built
files being created by the build (either upstream or downstream) or by
running automated tests, but not deleted. Those are easily fixed (several
fixed in git already).
libsdl2 has generated files in its upstream source tarball which are
re-generated with different content during the build (mostly Autotools
files, but also include/SDL_config.h and include/SDL_revision.h), and
I'm confident that it's not the only one: many upstreams want to do
this for the convenience of people building their package on OSs whose
development tools are less tightly integrated or less scriptable than
ours (especially Windows). In many cases the most pragmatic way to deal
with that is to delete the file during clean and ignore the resulting
warnings from dpkg-source.
I certainly don't want to require maintainers to invent machinery
for saving and restoring upstream's versions of generated files like
dh-autoreconf does, because that seems like busy-work that just adds to
the complexity of our builds without making Debian any better.
One way to streamline dealing with these generated files would be
to normalize repacking of upstream source releases to exclude them,
and make it easier to have source packages that genuinely only contain
what we consider to be source. At the moment, devref §6.8.8.2 strongly
discourages repacking tarballs to exclude DFSG-but-unnecessary files
(including generated files, as well as source/build files only needed on
Windows or macOS or whatever[1]), and Lintian strongly encourages adding
a +dfsg or +ds suffix to any repacked tarball, which makes it less
straightforward to track upstream's versioning. Is it time for us to
reconsider those recommendations?
For many upstreams (for example Autotools-based projects, and any project
like GTK that includes pre-generated documentation in source releases),
we can get "more source-like" upstream source releases by repacking our
own tarball based on upstream VCS tags than we would get by using their
official source release artifacts. For other upstreams, Files-Excluded
can be used to delete generated or unneeded files.
A side benefit of normalizing repacking upstream source releases would
be that maintainers are no longer expected to check the diff between
those files in the old and new version, and no longer required to track
the copyright and licensing status of the files that get excluded, which
can significantly speed up the process of importing new upstream versions
in some cases.
The major disadvantage of repacking upstream source is that if upstream
makes signed releases or publishes checksums, our repacked source cannot
be validated against that information; but if we can easily re-download
upstream source, then it would be possible for any interested developer
to verify that the repacked tarball matches what upstream released, minus
the files listed in Files-Excluded.
Devref §6.8.8.2 also says that "it is common for Debian users who
need to build software for non-Debian platforms to fetch the source
from a Debian mirror rather than trying to locate a canonical upstream
distribution point", but I'm not convinced that's true any more. Our
volunteers' time is our most limited resource, so if we can use that
time more efficiently by no longer catering to possibly-hypothetical
users who are building our source code on non-Debian platforms, then
that might be a worthwhile tradeoff.
smcv
[1] I mentioned Windows/macOS, but devref actually talks about MS-DOS,
which is perhaps an indication of how old that text is
Reply to: