1641414 - Intermittent racy failure of browser_resource_uri.js, especially in TV (test-verify) runs

Status: RESOLVED INCOMPLETE

(Core :: Internationalization, defect)

Description

[Randell Jesup [:jesup] (needinfo me)]

With bug 1602757 an apparent low/med-frequency intermittent when doing test-verify on browser_resource_uri.js becomes about a 50% fail (though it appears to rarely if ever fail outside of TV (test-verify) runs.

This bug predates bug 1602757; I can repro locally with a --verify run, and I tried it on a repo without process preallocation on, and it fails the same way occasionally.

The original patch for bug 1602757 caused this test to permafail on windows (only) during normal runs. Because of that, we added a patch to disable process preallocation for this test, but it still fails in test-verify. This also supports there being some type of race here.

The race manifests itself by document.l10n.ready getting rejected, via formatWithFallback(Sync?), with an error:
"[fluent] Request for keys failed because no resource bundles got generated.
1:02.10 GECKO(3155205) keys: [{"args":null,"id":"page-title"}].
1:02.10 GECKO(3155205) resourceIds: ["test.ftl"]."

zibi thinks this may be due to "it seems that intermittently either the Resource protocol has not added the new handler or the ChromeRegistry has not been settled when this is called"

Comment 1

[Zibi Braniecki [:zbraniecki][:gandalf]]

No failures in months