On Wed, 2012-11-07 at 04:55 -0500, Kamil Paral wrote:
> We get a *lot* of kernel bugs that are root-caused to be bad
memory
> or other hardware faults
> using the memtest on the install disc. Having it not be there by
> default is an extra step of
> hand-holding we need to do to get users to test things.
That is a good test case, thank you. And I agree memtest is useful. I
just have the feeling we should not hold the release when it's not
present (or, more concretely, I believe the criterion should not be
present, not that we shouldn't hold the release in specific cases. If
the problem arises, we should weight the situation and use common
sense. For example missing memtest is a smaller issue than memtest
reporting wrong results).
> Historically, how many times has this held up the release ?
I wouldn't like to go down this route when defining release criteria.
Worth considering the history: Petr proposed the general idea of a
criterion as part of the criteria review, and I drafted the final
wording. See thread "New criterion for Memory test" from end of January.
I don't _think_ I wrote it the way I did with the idea of sort of
'enforcing a design decision', though I certainly see that criticism. It
was more the idea that having the memory test actually is a key thing in
the Final build. But I think Kamil makes a reasonable argument that
maybe it isn't, I can certainly see the case for dropping the criterion.
It does seem unlikely that the memory test would somehow drop out as a
_quality_ issue; the only circumstance in which I could see that
happening was if we did a major rework on how the boot menu works, and
that would surely be a feature and get tested through the feature
process.
Peter, what do you think?
--
Adam Williamson
Fedora QA Community Monkey
IRC: adamw | Twitter: AdamW_Fedora | identi.ca: adamwfedora
http://www.happyassassin.net