search

shadow-maint / shadow

Upstream shadow tree
Other
292 stars 228 forks source link

4.14.x oldstable releases #926

Closed alejandro-colomar closed 3 months ago

alejandro-colomar commented 7 months ago

I'm wondering how much 4.14.x should live after 4.15.0. I'll keep here research about its users, and also use it to group issues about its releases.

(Edited to add more distros:)

Releases:

ikerexxe commented 7 months ago

Fedora 39 is using 4.14.0 and I don't plan to update to it, so you don't need to take it into account. Usually I only port patches that are requested by the community, and those requests don't happen very frequently.

alejandro-colomar commented 7 months ago

Then maybe I'll reduce the amount of patches that I backport to this series. I expect that build problems will not worry Alpine, or they would have complained already. Maybe just behavior bugs from now on (but 4.14.4 will still contain build fixes, since 4.15 won't be ready yet).

ikerexxe commented 7 months ago

I think that's the most sensible approach

jubalh commented 6 months ago

I believe there is a mistake on master and on the 4.14.4 release.

Our build system reports:

[   63s]  /usr/bin/install -c -m 644 chpasswd chfn chsh groupmems login newusers passwd su chage chpasswd groupadd groupdel groupmod useradd userdel usermod '/home/abuild/rpmbuild/BUILDROOT/shadow-4.14.4-203.1.x86_64/etc/pam.d'
[   63s] /usr/bin/install: will not overwrite just-created '/home/abuild/rpmbuild/BUILDROOT/shadow-4.14.4-203.1.x86_64/etc/pam.d/chpasswd' with 'chpasswd'

Reason most likely is here: https://github.com/shadow-maint/shadow/pull/928/files#r1487687347

chpasswd is present twice.

Since this is also on master I wonder whether we could guard against this in our CI?

alejandro-colomar commented 6 months ago

I believe there is a mistake on master and on the 4.14.4 release.

Thanks!

Our build system reports:

[   63s]  /usr/bin/install -c -m 644 chpasswd chfn chsh groupmems login newusers passwd su chage chpasswd groupadd groupdel groupmod useradd userdel usermod '/home/abuild/rpmbuild/BUILDROOT/shadow-4.14.4-203.1.x86_64/etc/pam.d'
[   63s] /usr/bin/install: will not overwrite just-created '/home/abuild/rpmbuild/BUILDROOT/shadow-4.14.4-203.1.x86_64/etc/pam.d/chpasswd' with 'chpasswd'

Reason most likely is here: https://github.com/shadow-maint/shadow/pull/928/files#r1487687347

chpasswd is present twice.

Hmm, yep, it looks like an accident. Cc: @DimStar77, @loqs, @dvzrv

Since this is also on master I wonder whether we could guard against this in our CI?

I would love to. Cc: @ikerexxe

DimStar77 commented 6 months ago

Since this is also on master I wonder whether we could guard against this in our CI?

I would love to. Cc: @ikerexxe

FWIW, this was not caught by the CI as it seems to build --without-libpam there. Only the combination --with-libpam --enable-account-tools-setuid would have triggered the CI to fail.

(the original reporter obviously tested --with-libpam --disable-account-tools-setuid, which was also fine)

To catch all cases, this would call for a rather complete matrix of configure parameters switching on/off

alejandro-colomar commented 6 months ago

FWIW, this was not caught by the CI as it seems to build --without-libpam there. Only the combination --with-libpam --enable-account-tools-setuid would have triggered the CI to fail.

(the original reporter obviously tested --with-libpam --disable-account-tools-setuid, which was also fine)

To catch all cases, this would call for a rather complete matrix of configure parameters switching on/off

Well, then I guess packagers act as a test of combinations actually in use. Not ideal, but not fatal.

Maybe we could test the combinations that downstreams are using.

ikerexxe commented 6 months ago

Since this is also on master I wonder whether we could guard against this in our CI?

I would love to. Cc: @ikerexxe

I don't think it's feasible to test all the configuration option combinations.

In any case, and from my perspective, it would be nice to include other distributions in our CI. Clearly we upstream maintainers cannot maintain all distributions, but if the downstream maintainers could do it, we would all win.

alejandro-colomar commented 5 months ago

There's another stable distro using 4.14.x:

alejandro-colomar commented 3 months ago

Fedora doesn't use this branch, Alpine hasn't upgraded since 4.14.2, and nixpkgs is EOL in a month.

I declare the 4.14.x branch EOL now, with its last release being 4.14.7 https://github.com/shadow-maint/shadow/releases/tag/4.14.7

alejandro-colomar commented 3 months ago

@hallyn , I can't remove the branch due to its protection. Please remove it.