Bug#935809: Re: buster-pu: package fdroidserver_1.1.4-1~deb10u1
Control: tags -1 - moreinfo
On Mon, 2019-08-26 at 23:37 +0200, Hans-Christoph Steiner wrote:
> Adam D. Barratt:
[...]
> > On Mon, 2019-08-26 at 14:54 +0200, Hans-Christoph Steiner wrote:
[...]
> They are mostly from compatibility bugs discovered since the final
> buster freeze. Some were fixed before the freeze, but we spent a lot
> of time testing this package on buster, so we wanted until the tests
> were complete before requesting an unblock. This release is meant to
> be an LTS for fdroidserver, and we'll be working to get it into other
> distros too.
[...]
> > > What
> > > makes the debdiff long is the cruft that were removed upstream in
> > > the
> > > source tarball due to a cleanup (see diffstat).
> >
> > Are you really saying that what looks from the diff like dropping
> > all localisation support is cruft cleanup? I feel like I may be
> > missing something.
>
> The localization was not removed. The localization user-editable .po
> files were removed while the .mo files are still there, which are
> what are actually used to display the strings.. The .mo files can be
> converted back to .po files as needed using msgunfmt, so they are
> also "source".
That seems an unusual approach, in that I would instead have expected
you to ship the .po files and generate the .mo at package build time,
but I guess it works.
> > In any case, cleaning up cruft by removing 60K lines of files from
> > your source package is generally not the sort of thing that would
> > be expected in a stable update.
>
> Yeah, this is annoying for sure. This cleanup was not triggered by
> me, it came from creating the source tarball on buster, while v1.1.1
> was created on stretch. The update to the standard Python tools that
> generate source tarballs seems to be the cause. I only noticed this
> after the release was done and uploaded. Sorry.
Well, you said above that you (as in upstream) intentionally removed
the PO files, so it seems you at least partly triggered it.
> > > The next largest block of changes is fixes to the test suite to
> > > work
> > > properly on Debian/buster.
> > > The remaining changes are compatibility fixes related to:
> > >
> > > 929905: incompatible with upstream's data repository
[...]
> > Please provide a diff that only includes the non-"cruft" changes,
> > so
> > there's a chance of us knowing what you're actually changing.
>
> I attached an excerpted debdiff without the aforementioned source
> tarball removals.
Thanks for that.
As a note, it may have been worth describing at least the "no-keep-git-
mirror-archive" and "archive-older" changes in your changelog(s). It's
possible they relate to one of the changes that you /have/ described,
but I'm honestly not sure.
+help=_("If a git mirror gets to big, allow the archive
to be deleted"))
That first "to" should be "too" - i.e. "gets too big".
Regards,
Adam
Bug#935809: Re: buster-pu: package fdroidserver_1.1.4-1~deb10u1
Adam D. Barratt: > Control: tags -1 + moreinfo > > Your report did not make it to debian-release due to the large size of > the attachment. I had to download the mbox from the BTS is order to > reply, which makes me quite surprised to see that you already decided > to upload the package. > > On Mon, 2019-08-26 at 14:54 +0200, Hans-Christoph Steiner wrote: >> I have been working on polishing this package for buster since March, >> but I didn't make the freeze cutoff. I'm part of upstream, and we >> did >> our Debian/buster bug fix work in our stable 1.1.x branch, so the >> changes are mostly in the upstream source tarball. >> >> The debdiff is long, but the actual code changes are quite short. > > If the changes are short, why were they not made in time for buster? They are mostly from compatibility bugs discovered since the final buster freeze. Some were fixed before the freeze, but we spent a lot of time testing this package on buster, so we wanted until the tests were complete before requesting an unblock. This release is meant to be an LTS for fdroidserver, and we'll be working to get it into other distros too. >> What >> makes the debdiff long is the cruft that were removed upstream in the >> source tarball due to a cleanup (see diffstat). > > Are you really saying that what looks from the diff like dropping all > localisation support is cruft cleanup? I feel like I may be missing > something. The localization was not removed. The localization user-editable .po files were removed while the .mo files are still there, which are what are actually used to display the strings.. The .mo files can be converted back to .po files as needed using msgunfmt, so they are also "source". > In any case, cleaning up cruft by removing 60K lines of files from your > source package is generally not the sort of thing that would be > expected in a stable update. Yeah, this is annoying for sure. This cleanup was not triggered by me, it came from creating the source tarball on buster, while v1.1.1 was created on stretch. The update to the standard Python tools that generate source tarballs seems to be the cause. I only noticed this after the release was done and uploaded. Sorry. >> The next largest block of changes is fixes to the test suite to work >> properly on Debian/buster. >> The remaining changes are compatibility fixes related to: >> >> 929905: incompatible with upstream's data repository > [...] >> ### 1.1.4 (2019-08-15) >> >> * include bitcoin validation regex required by fdroiddata > > As someone who knows nothing about fdroiddata, you may have to explain > why this is actually a problem. This is the sort of information that > should be in the request. To be honest, I'd have expected your upstream > diff and changelog to have included more information than that one > line. fdroiddata is like the collection of *.debian.tar.gz files for all the apps built by F-Droid. It is the collection of metadata, build instructions, and patches actually used to run the builds. For anyone who wants to contribute work related to apps in F-Droid, they need to use fdroidserver. These changes make buster's version work with all of the files currently in the fdroiddata collection (its about 3000 apps currently). > Please provide a diff that only includes the non-"cruft" changes, so > there's a chance of us knowing what you're actually changing. I attached an excerpted debdiff without the aforementioned source tarball removals. Here's the diff to show which files were removed in this debdiff excerpt: --- fdroidserver_1.1.1-1_1.1.4-1.debdiff +++ fdroidserver_1.1.1-1_1.1.4-1~deb10u1_excerpt.debdiff - .gitignore | 59 - .gitlab-ci.yml | 198 -- - .pylint-rcfile | 45 - .travis.yml| 109 - CHANGELOG.md | 16 - MANIFEST.in| 550 ++ PKG-INFO |6 - buildserver/.gitignore |4 completion/bash-completion |4 debian/changelog | 31 debian/control |2 fdroidserver.egg-info/PKG-INFO |6 fdroidserver.egg-info/SOURCES.txt | 38 fdroidserver.egg-info/requires.txt | 20 fdroidserver/index.py |1 fdroidserver/metadata.py |2 fdroidserver/nightly.py|8 fdroidserver/publish.py|2 fdroidserver/scanner.py|1 fdroidserver/server.py |6 - hooks/install-hooks.sh | 30 - hooks/pre-commit | 136 - - jenkins-build-all | 107 - - jenkins-setup-build-environment| 119 - - jenkins-test | 78 -
Bug#935809: Re: buster-pu: package fdroidserver_1.1.4-1~deb10u1
Control: tags -1 + moreinfo Your report did not make it to debian-release due to the large size of the attachment. I had to download the mbox from the BTS is order to reply, which makes me quite surprised to see that you already decided to upload the package. On Mon, 2019-08-26 at 14:54 +0200, Hans-Christoph Steiner wrote: > I have been working on polishing this package for buster since March, > but I didn't make the freeze cutoff. I'm part of upstream, and we > did > our Debian/buster bug fix work in our stable 1.1.x branch, so the > changes are mostly in the upstream source tarball. > > The debdiff is long, but the actual code changes are quite short. If the changes are short, why were they not made in time for buster? > What > makes the debdiff long is the cruft that were removed upstream in the > source tarball due to a cleanup (see diffstat). Are you really saying that what looks from the diff like dropping all localisation support is cruft cleanup? I feel like I may be missing something. In any case, cleaning up cruft by removing 60K lines of files from your source package is generally not the sort of thing that would be expected in a stable update. > The next largest block of changes is fixes to the test suite to work > properly on Debian/buster. > The remaining changes are compatibility fixes related to: > > 929905: incompatible with upstream's data repository [...] > ### 1.1.4 (2019-08-15) > > * include bitcoin validation regex required by fdroiddata As someone who knows nothing about fdroiddata, you may have to explain why this is actually a problem. This is the sort of information that should be in the request. To be honest, I'd have expected your upstream diff and changelog to have included more information than that one line. Please provide a diff that only includes the non-"cruft" changes, so there's a chance of us knowing what you're actually changing. Regards, Adam
