Bug 134918
| Summary: | DRT/WTR should clear the memory cache between tests | ||
|---|---|---|---|
| Product: | WebKit | Reporter: | Simon Fraser (smfr) <simon.fraser> |
| Component: | Tools / Tests | Assignee: | Antti Koivisto <koivisto> |
| Status: | NEW | ||
| Severity: | Normal | CC: | ap, cdumez, dino, kling, koivisto, simon.fraser |
| Priority: | P2 | ||
| Version: | 528+ (Nightly build) | ||
| Hardware: | Unspecified | ||
| OS: | Unspecified | ||
Simon Fraser (smfr)
DRT doesn't seem to clear the memory cache between tests. I was trying to make a test that has window.internals.settings.setImageSubsamplingEnabled(true); and a ref test that does not, but the CachedImage was created by the test then just re-used by the reference, giving me a bogus pass.
| Attachments | ||
|---|---|---|
| Add attachment proposed patch, testcase, etc. |
Simon Fraser (smfr)
Antti says:
I think that is a too big of a hammer. You can use internals.clearMemoryCache() if your test relies on empty memory cache.
Chris Dumez
I am personally in favor of this change. I will help reduce flakiness. I had to add internals.clearMemoryCache() calls at the beginning of 24 tests recently to make them not flaky, I don't think this is very scalable. We already disable PageCache during testing, I think it makes sense to make the MemoryCache opt-in as well.
Chris Dumez
(In reply to comment #2)
> I am personally in favor of this change. I will help reduce flakiness. I had
> to add internals.clearMemoryCache() calls at the beginning of 24 tests
> recently to make them not flaky, I don't think this is very scalable. We
> already disable PageCache during testing, I think it makes sense to make the
> MemoryCache opt-in as well.
I would also add that I wasted engineering time investigating the source of the flakiness until I was able to determine the reason was the memory cache. This WILL happen again and more people WILL waste time on this.
Simon Fraser (smfr)
Reopening.
Alexey Proskuryakov
One concern if that *may* make the tests substantially slower. We'll need to measure performance if we do this.
Antti Koivisto
internals.clearMemoryCache is not to meant to be flakiness fix but used when a test actually relies on having a clear cache.
Doing it everywhere would slow down tests (no idea if enough to matter) and reduce implicit test coverage, moving testing further from real browser behavior in general.