search

canonical / ubuntu-mir

Ubuntu Main Inclusion Process - formerly on https://wiki.ubuntu.com/MainInclusionProcess
14 stars 11 forks source link

Do we benefit from having the dev binary packages promoted? #69

Open seb128 opened 1 week ago

seb128 commented 1 week ago

I'm opening one to start a follow discussion to the 0. review comment on https://bugs.launchpad.net/bugs/2080965

0 - This is mostly a build-time helper tool, do we really need it in main, or should we consider demoting reverse-depends (like libglib2.0-dev) to universe instead?

Demoting libglib2.0-dev isn't easy because it's a low level libraries that most other desktop dev libraries depends on. We regularly have issues with dev packages and end up demoting and excluding some of those. Which makes me wonder, why do we default to promote those? Should we reconsider that position?

slyon commented 1 week ago

By default, we promote packages on a source package level, only excluding specific binaries (such as *-dev packages) as needed. Also, we do not generally require Build-Depends to be in "main", only runtime depends.

The important exception here is for build-time tooling that affects the resulting binaries (e.g. compilers), and other critical infrastructure that we depend on in our daily work as Ubuntu developers.

During MIR review we already watch out for cases where we'd want to exclude -dev packages, if they are not strictly necessary and bring unnecessary extra work (e.g. pulling fonts packages into main from a -doc package). The template reads:

- no -dev/-debug/-doc packages that need exclusion

Such cases would be handled using Extra-Exclude in the seeds.

slyon commented 1 week ago

This was discussed during today's MIR meeting.

https://ubottu.com/meetingology/logs/ubuntu-meeting/2024/ubuntu-meeting.2024-09-24-14.31.log.html (starting around the 14:44 timestamp).

There wasn't a clear consensus for changing the default -dev auto-inclusion. We should probably work it out on a case-by-case basis instead. libglib2.0-dev is probably a special case here, due to its long-tail of reverse-depends.