Bug 248824 – Many cache-storage tests failing on EWS in sequence

• Status: RESOLVED
• Resolution: FIXED
• Component: Service Workers

[Jonathan Bedard]

EWS results indicate it's possible for cache-storage tests to kick a worker into a bad state, causing all cache-storage tests in that worker to fail. Some examples of test runs where we're seeing this:
https://ews-build.webkit.org/#/builders/70/builds/23009
https://ews-build.webkit.org/#/builders/70/builds/23008
https://ews-build.webkit.org/#/builders/70/builds/22996
https://ews-build.webkit.org/#/builders/70/builds/22992

[Radar WebKit Bug Importer]

<rdar://problem/103030498>

[Jonathan Bedard]

Pull request: https://github.com/WebKit/WebKit/pull/7209

[EWS]

Test gardening commit 257415@main (9d139ae059b9): <https://commits.webkit.org/257415@main>

Reviewed commits have been landed. Closing PR #7209 and removing active labels.

[Ryan Haddad]

Pull request: https://github.com/WebKit/WebKit/pull/7212

[EWS]

Test gardening commit 257416@main (d53145462ffa): <https://commits.webkit.org/257416@main>

Reviewed commits have been landed. Closing PR #7212 and removing active labels.

[Jonathan Bedard]

The first test we see this in is https://ews-build.webkit.org/#/builders/70/builds/22896.

With that in mind, I think the "regression" range is:
http://commits.webkit.org/compare/257364@main...257373@main

That said, I put regression in quotes, because I suspect https://commits.webkit.org/257367@main changed test ordering to "cause" this, which doesn't really help us narrow down the issue.

[Jonathan Bedard]

Pull request: https://github.com/WebKit/WebKit/pull/7220

[Jonathan Bedard]

https://ews-build.webkit.org/#/builders/70/builds/23167 is when we skip the css-nesting tests
https://ews-build.webkit.org/#/builders/70/builds/23212 is when we run the css-nesting and our cache-storage tests.

https://commits.webkit.org/257367@main is definitely the point where EWS regressed, but doesn't really help us determine when the behavior itself regressed, and perhaps it isn't even a regression, but rather, a problem our cache-storage tests have always had and was just revealed by changing test order slightly.

[youenn fablet]

(In reply to Jonathan Bedard from comment #8)
> https://ews-build.webkit.org/#/builders/70/builds/23167 is when we skip the
> css-nesting tests
> https://ews-build.webkit.org/#/builders/70/builds/23212 is when we run the
> css-nesting and our cache-storage tests.
> 
> https://commits.webkit.org/257367@main is definitely the point where EWS
> regressed, but doesn't really help us determine when the behavior itself
> regressed, and perhaps it isn't even a regression, but rather, a problem our
> cache-storage tests have always had and was just revealed by changing test
> order slightly.

The cache-storage tests seem to pass ok in https://ews-build.webkit.org/#/builders/70/builds/23212, I only see a gamepad/gamepad-polling-access.html failure.
I do not see a proof that would tell us whether test order is the reason, nor do we have a proof to rule it out.

https://ews-build.webkit.org/#/builders/70/builds/2321 seems to indicate we could try reenable cache-storage tests.

[Sihui Liu]

Pull request: https://github.com/WebKit/WebKit/pull/12207

[EWS]

Committed 262380@main (9579456a8bc1): <https://commits.webkit.org/262380@main>

Reviewed commits have been landed. Closing PR #12207 and removing active labels.