> Of the `enabled-by-*` geolocation tests, in comment 8 I had managed to choose the only one that isn't fixed by setting > `geo.prompt.open_system_prefs=false` That was wrong. They still had `geo.prompt.testing`. I may have found the main issue and I should be able to draft a patch for it. There are few problems and some outstanding mysteries. First, `geo.prompt.open_system_prefs=false` should of course be set for all tests -- that was [done in bug 1900225 for mochitests](https://searchfox.org/mozilla-central/rev/70fb2d2a702db4cb455d282be77ecfe9bdcea589/testing/profiles/mochitest/user.js#40-43) but not for wpt. That's the easy part. After that, the `enabled-by-permission-policy-attribute*` tests still timeout due to a dialog, but this time its the usual Firefox geolocation permission one. The issue here seems to be that [a commit went into wpt in June that changed the test](https://github.com/mozilla/gecko-dev/commit/f1abdf0d5a34f62575e045d37d3a3a24073701db#diff-ccf7ee088cb5483fa337ee762af1d6adb9695b8549e554c61d4809c62a49e0a2) -- the checkin comment suggests it's going to move the place where the top-level frame is granted permission (these are iframe inherited permission tests -- the "attribute" from the test filename is the "allow" attribute on the iframe). Instead, it just deletes it. So the top level didn't have permission so the iframes had nothing to inherit. Adding the `set_permission` call to the test files as suggested does work... although there is ALSO an intermittent issue, so it seems to only pass about half the time for me locally. I am utterly baffled at how we are passing these in CI, especially since [wpt.fyi](https://wpt.fyi/results/geolocation?label=master&label=experimental&product=firefox-130.0a1&aligned) seems to believe that no browsers have been passing these tests for many months. (Note that that link is for results from last August.) Worse, I saw the same pattern for a bunch of other wpt permissions tests, such as the battery API. All of those tests also fail for me locally.
Bug 1937188 Comment 12 Edit History
Note: The actual edited comment in the bug view page will always show the original commenter’s name and original timestamp.
> Of the `enabled-by-*` geolocation tests, in comment 8 I had managed to choose the only one that isn't fixed by setting > `geo.prompt.open_system_prefs=false` That was wrong. They still had `geo.prompt.testing`. I may have found the main issue and I should be able to draft a patch for it. There are few problems and some outstanding mysteries. First, `geo.prompt.open_system_prefs=false` should of course be set for all tests -- that was [done in bug 1900225 for mochitests](https://searchfox.org/mozilla-central/rev/70fb2d2a702db4cb455d282be77ecfe9bdcea589/testing/profiles/mochitest/user.js#40-43) but not for wpt. That's the easy part. After that, the `enabled-by-permission-policy-attribute*` tests still timeout due to a dialog, but this time its the usual Firefox geolocation permission one. The issue here seems to be that [a commit went into wpt in June that changed the test](https://github.com/mozilla/gecko-dev/commit/f1abdf0d5a34f62575e045d37d3a3a24073701db#diff-ccf7ee088cb5483fa337ee762af1d6adb9695b8549e554c61d4809c62a49e0a2) -- the checkin comment suggests it's going to move the place where the top-level frame is granted permission (these are iframe inherited permission tests -- the "attribute" from the test filename is the "allow" attribute on the iframe). Instead, it just deletes it. So the top level didn't have permission so the iframes had nothing to inherit. Adding the `set_permission` call to the test files as suggested does work... although there is ALSO an intermittent issue, so it seems to only pass about half the time for me locally. I am utterly baffled at how we are passing these in CI, especially since [wpt.fyi](https://wpt.fyi/results/geolocation?label=master&label=experimental&product=firefox-130.0a1&aligned) seems to believe that no browsers have been passing these tests for many months. (Note that that link is for results from last August.) Worse, I saw the same pattern for a bunch of other wpt permissions tests, such as [the battery API](https://wpt.fyi/results/battery-status?label=master&label=experimental&product=firefox-130.0a1&aligned). All of those tests also fail for me locally.