New issue
Advanced search Search tips

Issue 805225 link

Starred by 2 users
Status: Available
Owner: ----
Cc:
falken@chromium.org
leon....@intel.com
shimazu@chromium.org
Components:
EstimatedDays: ----
NextAction: ----
OS: ----
Pri: 2
Type: Bug

Blocking:
issue 678905



Sign in to add a comment

Pass WPT: unregister-then-register-new-script.https.html

Project Member Reported by falken@chromium.org, Jan 24 2018 
Mozilla recently changed the expectation of the test regarding resurrection to match the spec. Apparently it was not what the spec was saying before.

From https://bugzilla.mozilla.org/show_bug.cgi?id=1425975#c20:
"This fixes a bad WPT test.  It was incorrectly trying to assert that registering a 404 or reject-on-install worker would not resurrect an uninstalling registration.  There is no such spec language, AFAICT.  See step 5.1 here:

https://w3c.github.io/ServiceWorker/#register-algorithm

It clears the flag before running the update algorithm with does the fetching and install event dispatch."

Now we fail with:
FAIL Registering a new script URL that 404s does resurrect an unregistered registration assert_unreached: unexpected rejection: Cannot read property 'scriptURL' of null Reached unreachable code
FAIL Registering a new script URL that fails to install does resurrect an unregistered registration assert_unreached: unexpected rejection: Cannot read property 'scriptURL' of null Reached unreachable code

Comment 1 by bke...@mozilla.com, Jan 24 2018 

FWIW I also added a pointer to the relevant spec text in the test:

https://github.com/w3c/web-platform-tests/blob/master/service-workers/service-worker/unregister-then-register-new-script.https.html#L148

If this is in the spec somewhere I'd be happy to revisit it.

Comment 2 by falken@chromium.org, Jan 24 2018 

Thanks Ben, I think this change was right (the test is very old and was probably written before a spec or for an old spec).

Comment 3 by minggang...@intel.com, Nov 1 

Based on the latest decision from https://github.com/w3c/ServiceWorker/issues/1204 about the behavior of resurrection, I suppose we should re-verify this case when the test case is updated. I have been investigating this issue for a couple of days, and understand some internal implementations inside the blink, so I'd like to contribute to it anytime if needed, thanks!

Comment 4 by leon....@intel.com, Nov 2

Cc: falken@chromium.org shimazu@chromium.org
+falken@, shimazu@
Seems we need to wait for the redefinition of the behavior when calling unregister() on a registration with an active worker that is still controlling clients, right? Then we'd fix our impl based on the new spec/tests.

Comment 5 by leon....@intel.com, Nov 2

Cc: leon....@intel.com

Comment 6 by falken@chromium.org, Nov 2 

Agree, best to wait for resurrection to be removed.

Comment 7 by minggang...@intel.com, Nov 2 

I am willing to verify it when the removal is done and fix any problems.
►

Sign in to add a comment