Sam James <[email protected]> writes:

> Arthur Zamarin <[email protected]> writes:
>
>> Signed-off-by: Arthur Zamarin <[email protected]>
>> ---
>>  ...25-08-23-stable-hppa-keywords-removed.en.txt | 17 +++++++++++++++++
>>  1 file changed, 17 insertions(+)
>>  create mode 100644 
>> 2025-08-23-stable-hppa-keywords-removed/2025-08-23-stable-hppa-keywords-removed.en.txt
>
> LGTM. I'm of course a little sad over it, but it's too expensive for me
> to run my own machine 24/7, and the backlog of trivial stablereqs means
> then I don't have much time/energy to work on "real" hppa bugs.
>
> So, if anything, this (and the sparc change) may well mean the arch ends
> up in a better state.

I was going to write an explanation for somebody else and figured I'd
write it here just for completeness: the reason we're doing this (and
for hppa) is that stabilisation requires a high amount of effort.

The following is so I can link to it when people ask..

Many packages get frequent releases. We recommend 30 days before a
new version is considered for stabilisation (exceptions apply for
special packages where they need updates to keep working with an
external service, security fixes, minor bug fix releases for
regressions, etc). If something starts to fail tests on an arch, this
blocks the process and someone has to investigate.

This of course is invaluable in finding regressions, but it requires a
lot of time to sit and check what an issue was. On slower machines,
they're often timeouts or some other transient failure. But you can't
know that until you sit down and check it. And there's a huge number of
other such issues which affect arches/platforms where stabilisation
actually benefits people.

sparc and hppa (in particular) never had huge userbases but the question
is not "is stable a good thing", but rather "is it actually helping
people on this arch, and is it worth it"?

Combined with the lack of hardware right now (our hppa machine seems to
have died and my home one is too expensive to run 24/7; our sparc one
uses too much power and the power management patches from Oracle were
never ported to newer Linux versions or upstreamed), all we're really
doing is making people frustrated with our portability because the
delays in sparc or hppa completing stabilisation hold up the whole bug,
cleanups, and also prevent automation from flagging new versions for
stabilisation for other arches.

Destabilisation doesn't mean we're going to remove an arch, just that
the trade-off for being able to compile & run tests for a package very
often (and having someone be able to check on failures) has moved for
these arches.

sam

> [...]

Reply via email to