1180746 - "Harness status: OK" + failing test when running "sync-xhr-doesnt-deadlock.https.html" test

Status: RESOLVED DUPLICATE of bug 931243

(Testing :: web-platform-tests, defect)

Description

[Noemí Freire (:noemi)]

Checked with 7/6 master build

Test run such as |./mach web-platform-tests _mozilla/service-workers/service-worker/sync-xhr-doesnt-deadlock.https.html|

Result:
* Harness status: OK
* Found 1 test
* 1 Fail:
**Verify SyncXHR does not deadlock	
***assert_unreached: unexpected rejection: FAIL Reached unreachable code
unreached_rejection/<@https://web-platform.test:8443/_mozilla/service-workers/service-worker/resources/test-helpers.sub.js:42:7 Test.prototype.step@https://web-platform.test:8443/resources/testharness.js:1363:20 Test.prototype.step_func/<@https://web-platform.test:8443/resources/testharness.js:1387:1 Promise*@https://web-platform.test:8443/_mozilla/service-workers/service-worker/sync-xhr-doesnt-deadlock.https.html:12:5 Test.prototype.step@https://web-platform.test:8443/resources/testharness.js:1363:20 async_test@https://web-platform.test:8443/resources/testharness.js:513:13 @https://web-platform.test:8443/_mozilla/service-workers/service-worker/sync-xhr-doesnt-deadlock.https.html:8:1 
* Traces: https://pastebin.mozilla.org/8838618

Comment 1

[Nikhil Marathe [:nsm] (No longer reading bugmail, please needinfo?)]

It seems like Blink had some issue where sync XHR and fetch interception could deadlock. Is this something we need to worry about? From my reading of our sync xhr implementation, we continue to spin an event loop (albeit a nested event loop), so runnables coming back from the service worker will run, and rendering should also continue to happen. Am I right? In that case this test is irrelevant and probably should be removed from wpt tests.

Olli, you were the last person to touch the sync loop creation in XHR. Any thoughts?

[1] https://codereview.chromium.org/553333005/

Comment 2

[Olli Pettay [:smaug][[e-mail]]]

Our main thread sync XHR just spins the event loop (not really nested, I'd say, just the same one on which the XHR is running. But depends on how one defines "nested").
Rendering does happen and runnables are processed normally. But note, our behavior on the main thread is not right - we shouldn't really do anything while waiting for the result for sync XHR.

Can one use sync XHR on the service worker? If so, can it cause problems?

Comment 3

[Andrew Overholt [:overholt]]

This may be a non-issue since in bug 931243 we're going to disable sync XHR in Service Workers.

Comment 4

[(no longer active)]

My patch for bug 931243 removes this test.