[11:25] <seb128> tjaalton, hey, isn't https://salsa.debian.org/xorg-team/lib/libxcvt/-/blob/debian-unstable/debian/tests/control# testing the src rebuilt rather than the actually archive binary?
[11:40] <tjaalton> seb128: how so?
[11:42] <tjaalton> it was copied from some other package as suggested by someone
[11:49] <seb128> tjaalton, well, for one thing that package seems to have no tests?
[11:49] <seb128> https://autopkgtest.ubuntu.com/results/autopkgtest-kinetic/kinetic/amd64/libx/libxcvt/20220327_202617_8878e@/log.gz
[11:49] <seb128> 	cd obj-x86_64-linux-gnu && LC_ALL=C.UTF-8 MESON_TESTTHREADS=1 meson test
[11:49] <seb128> No tests defined.
[11:50] <tjaalton> upstream has no build tests, true
[11:50] <tjaalton> did you look at tests/build-test?
[11:50] <seb128> tjaalton, shrug, copy paste mangled the URL I wanted to share
[11:51] <seb128> tjaalton, it was meant to be https://salsa.debian.org/xorg-team/lib/libxcvt/-/blob/debian-unstable/debian/tests/control#L4-6
[11:52] <tjaalton> ah, well that part could be dropped I guess
[11:53] <seb128> tjaalton, also in principle even if you had build tests, those usually test what you just built, not the system version
[11:53] <tjaalton> sure
[11:54] <tjaalton> why is that a problem?
[11:54] <seb128> because imagine you have an ABI change in one of your depends, like different objects order in a struct
[11:54] <seb128> your library would be buggy until rebuilt
[11:54] <seb128> your test is against the rebuilt, which wouldn't see the issue
[11:55] <seb128> but then migrating the depends would break libxcvt in the archive
[11:55] <seb128> until it's rebuilt
[11:55] <tjaalton> well the other test would catch that
[11:55] <seb128> why would it?
[11:55] <seb128> it's a built test
[11:55] <tjaalton> it builds a test app against the archive version
[11:55] <tjaalton> aiui
[11:55] <seb128> yes
[11:56] <seb128> but like in the struct definition case
[11:56] <seb128> a rebuild would pick the new definition from the depends
[11:56] <seb128> and the build + test would work
[11:56] <seb128> which tells you that a rebuild of libxcvt would work fine
[11:56] <seb128> but doesn't tell you that $depends shouldn't migrate until libxcvt it rebuilt because the current deb is impacted by the ABI change
[11:57] <tjaalton> is this for an actual case or just for a theoretical issue? libxcvt has no depends other than libc6
[11:58] <seb128> theoretical, I'm just pointing out that the approach is wrong
[11:58] <seb128> the reason I'm mentioning it is because nteodosio copied that for adding a test to webp-pixbuf-loader
[11:58] <tjaalton> then I bet there are plenty of other packages which are wrong
[11:58] <seb128> right, which is why I'm pointing it out
[11:59] <seb128> so we stop copying a buggy case to new packages
[11:59] <seb128> sorry if I went the wrong way
[11:59] <tjaalton> I haven't found a "best practices with simple linking examples for autopkgtests", other than blindly trying to look for examples from other packages
[12:00] <tjaalton> I don't remember where this was copied from, and who gave the idea
[12:00] <tjaalton> probably when the MIR was being reviewed
[12:02] <tjaalton> but still, I don't see how building the test app would "rebuild" anything
[12:08] <slyon> enr0n: here are some of my remarks about the systemd 251 merge: https://code.launchpad.net/~enr0n/ubuntu/+source/systemd/+git/systemd/+merge/425171/comments/1125387 – How do you feel about the overall state of the package (especially wrt to autopkgtests)? Do you feel like we can upload this later this week?
[12:55] <enr0n> slyon: thanks for reviewing! I do feel lke we can upload later this week. WRT autopkgtests, I would like to continue to improve the upstream-{1,2} test consistency over the course of the cycle (already on the roadmap). Besides that, I hope that we can address a few autopkgtest quirks I had to make (e.g. skipping CoredumpTest on s390x), but I don't think that should hold up the upload.
[12:57] <slyon> enr0n: I agree. sounds good to me!
[12:57] <slyon> I'll happily sponsor that for you, maybe after we got some of my comments addressed.
[13:01] <seb128> tjaalton, sorry had a meeting, you are right that build-app test doesn't have an issue, my point was mostly about the dh_auto_test one
[13:12] <enr0n> slyon: perfect, I will work on addressing your comments this morning
[13:47] <tjaalton> seb128: right, and it's dropped in git
[13:47] <seb128> tjaalton, thanks!
[13:48] <seb128> tjaalton, and sorry if I was unclear in what I was trying to get at before
[13:48] <tjaalton> nw
[14:02] <schopin> Does switching a library from mbedTLS to openssl break its ABI? In this case, libgit2.
[14:09] <edugomezescandel> Hi all, I sumbitted a Merge Request to the ubuntu-release-upgrader, can anyone give it a look?
[14:09] <edugomezescandel> It's this one: https://code.launchpad.net/~edugomez/ubuntu/+source/ubuntu-release-upgrader/+git/ubuntu-release-upgrader/+merge/425145
[14:09] <edugomezescandel> It's just a bugfix for WSL
[15:30] <bdmurray> edugomezescandel: I've seen it and will try and get to it. Are you also looking to see this fixed for 22.04?
[15:33] <edugomezescandel> Hi, thanks! Yes, that would be the idea.