Closed Bug 1005936 Opened 10 years ago Closed 5 years ago

Machine sanity for releng infra

Categories

(Testing :: General, defect)

x86_64
Linux
defect
Not set
normal

Tracking

(Not tracked)

RESOLVED INACTIVE

People

(Reporter: armenzg, Unassigned)

References

Details

We should not loose this ability as it undermines development.
sheriffs: what are your thoughts about this proposal?

I think we could have a very small suite that would check that releng infra is setup the way that we need it:
* If it crashes, it creates a dump
* We check on the minimum screen resolution
** We did not have the right screen resolution for more than a year on Win7 and did not run accelerated reftests
* The screen saver is disabled

Any other useful tests?

We could run this nightly.
Summary: Create test to crash the browser and make sure that we get a crash dump → Create test to catch certain expected behaviours from the releng infra
We already have automated tests for crash handling, so I'm not sure that we need extra tests for that case.
(In reply to Benjamin Smedberg  [:bsmedberg] from comment #2)
> We already have automated tests for crash handling, so I'm not sure that we
> need extra tests for that case.

What are the required components need in place for a crash dump to happen?
An environment variable? The path to the breakpad binaries? [1]

I'm trying to make sure that we don't loose such ability on the releng infra.
AFAIK, right now, we have to *hope* that it will stay working.

If you believe it is unnecessary feel free to mention it and the why. We can change the scope if so.

[1] http://hg.mozilla.org/build/tools/file/cd95bd41bb5c/breakpad/win32
> jmaher: armenzg: it would be great if we detected leftover files on the system as well :)

If we create the tests, we can ask releng to run it pre-buildbot.

Meanwhile, if we make the test run fast we might be able to run it on every job before running the Fx tests et al.

This way we can catch bad slaves much much faster.
Blocks: 712206
Summary: Create test to catch certain expected behaviours from the releng infra → Machine sanity for releng infra
Yeah, thinking it over more, I think getting this run pre-flight on each slave prior to taking a test job is probably the better option (more deterministic failures when a slave is misconfigured and after a much shorter period of time). But in principle, this sounds great to me!
No longer blocks: 712206
Depends on: 712206
Status: NEW → RESOLVED
Closed: 5 years ago
Resolution: --- → INACTIVE
You need to log in before you can comment on or make changes to this bug.