Question on procedure

[Brian Dessent]

NightStrike wrote:

> If a testsuite fails for a target, should I write a new PR for it, or
> just send an email to gcc-bug?  Also, should I do so for all
> unexpected failures?

The gcc-bugs mailing list is really just for the output of bugzilla
generated traffic.  Bugs not reported through bugzilla are not tracked,
so you'd just be wasting your breath, as certainly not many people would
even see it and those that do would ignore it or only reply to tell you
to use bugzilla.  I'm not even sure why gcc-bugs@ is not set to
read-only like gcc-cvs@, given that someone sends such a fruitless
message to it every week or two or so.

And I certainly wouldn't spam a bunch of target-specific PRs for failing
test cases either.  If you look at the gcc-testresults@ archives you'll
see that nearly every target has dozens to hundreds of failing tests, so
expecting to reach let alone maintain 0 is just not being realistic,
especially during phase 1.  If every target opened a PR for every
failing test then bugzilla would be swamped with useless noise.

IMHO you should open a PR when a testcase that previously passed starts
to fail as a result of a recent checkin that you can identify, or when
you can identify a flaw in a testcase (i.e. it's an invalid test.)  But
just reporting that a testcase is currently failing with no additional
information -- especially on an embryonic target -- would be useless
bugspam, IMHO.

Brian