test262.report webpage shows 504 Gateway Time-out error

[Constellation]

Summary

test262.report shows 504 Gateway Time-out error.

Expected behavior

Showing the page.

Actual behavior

test262.report shows 504 Gateway Time-out error

Steps to reproduce

1. Open test262.report in a browser

Environment

macOS Safari 14

[codehag]

@boazsender @isaacdurazo is there any news about this? Quite a few of us are surprised to see the site down. Anything we can do to help?

[jugglinmike]

I'm hoping to have this resolved by early next week. I'll keep you folks posted!

[jugglinmike]

Status update: Our automated testing system can no longer collect results from
JavaScriptCore. Following a recent release, the process crashes during Intl
tests due to memory exhaustion. I've been able to reproduce that behavior on my
GNU/Linux development machine using JSVU and test262-harness, so I don't
believe the problem is isolated to our test environment.

The next step is to verify the presumed regression by bisecting releases of
JSC. I've been told this is possible for someone external to Apple, provided
they have a MacOS machine, via
https://webkit.org/blog/7978/introducing-webkit-build-archives/ ’s
OpenSource/Tools/Scripts/bisect-builds.

I don't yet know when I can begin that research, but I will update this issue
when I do.

[ljharb]

@jugglinmike could we disable JSC results entirely for the purpose of getting the website up sooner?

[jugglinmike]

@ljharb At this moment, it's not clear whether that would be a faster course of action.

[ljharb]

I guess the other question is; why would running tests impact serving the website? Aren’t those entirely separate processes?

[Constellation]

@jugglinmike What test causes the failure? WebKit is maintaining CI running test262, and so far, we are not seeing that failure.
https://build.webkit.org/#/builders/8

[jugglinmike]

It appears to be due to system resource exhaustion. Safari's memory utilization seems to have increased beyond the capacity of the CI environment. We're seeing promising results after reducing the amount of parallelization used to collect results from that engine, though!