Tuesday, 9 April 2019

Re: Proposal: drop rails and its reverse-dependencies from Ubuntu 19.04 [Re: Maintaining language-specific module package stacks]


On 4/9/19 10:15 PM, Matthias Klose wrote:
> On 10.04.19 04:37, Simon Quigley wrote:
>> On 4/9/19 9:27 PM, Matthias Klose wrote:
>>> rails is ready to migrate, there is no puma package in the release pocket. the
>>> failing puma autopkg test in -proposed shouldn't be any concern.
>>> Filed LP: #1824049 for that.
>>> Now we could go on removing puma from -proposed, and then rails should migrate.
>>> How can we do that without removal?
>> (disclaimer: not on the release team)
>> This isn't a bug in Britney; Britney is designed to block on *any*
>> autopkgtest failures if there aren't any test hints (thus, a documented
>> reason for it failing). Passing autopkgtests for all packages is a
>> release goal, and unless the package has a hint (which is an exception
>> to the rule), any failing autopkgtests shouldn't let a package into the
>> release pocket. This autopkgtest should be evaluated to see if it's a
>> real regression in rails or if it's puma autopkgtests not working properly.
> Call it a britney bug or a proposed-migration bug. But to what extent should we
> care about a regression which doesn't show in the software that we ship? It's
> certainly not contradicting your statement "Passing autopkgtests for all
> packages is a release goal", because puma then wouldn't be part of the release.
> Now remove rails and dependencies and you might be able to update to a new ruby
> version much earlier, with even more regressions outside the archive.

I think we should care about it to the extent that we can ensure that
puma's failure is not caused by ruby. If puma's autopkgtests are
flaky/unfixable without Very Horrible Hacks, they should be hinted. If
the failures are irrelevant to the ruby update, ruby should be hinted
when all other rdep pass their tests. If the puma test failure is easy
to fix, we should solve that, and then we're done here. All of these
actions can be completed without the Archive Admins.

I still don't personally see this as a bug in currently-deployed code.

Simon Quigley
tsimonq2 on freenode and OFTC
5C7A BEA2 0F86 3045 9CC8
C8B5 E27F 2CF8 458C 2FA4