fix: destroy WebContents synchronously on shutdown #15541

zcbenz · 2018-11-02T05:30:18Z

Description of Change

Destroying WebContents asynchronously on shutdown may delay the destroying task after the shutdown task, resulting in an assertion in ResourceDispatcherHostImpl::OnShutdown.

This PR destroys WebContents synchronously on shutdown, so we can fix #15133 without breaking #8930.

I have also added notes for DestroyWebContents on how to use it correctly.

Checklist

PR description included and stakeholders cc'd
npm test passes
tests are changed or added
relevant documentation is changed or added
PR title follows semantic commit guidelines

Release Notes

Notes: Fix crash on exit when using BrowserView.

zcbenz · 2018-11-02T10:23:13Z

This PR has accidentally revealed a crash in other place:

* thread #1: tid = 0x83cc89, 0x000000010122ee88 Electron Framework`::GuestCurrentlyAudibleCallback() + 8 at browser_plugin_embedder.cc:200, name = 'CrBrowserMain', queue = 'com.apple.main-thread', stop reason = EXC_BAD_ACCESS (code=1, address=0x190)
  * frame #0: 0x000000010122ee88 Electron Framework`::GuestCurrentlyAudibleCallback() + 8 at browser_plugin_embedder.cc:200 [opt]
    frame #1: 0x00000001033de9c1 Electron Framework`::ForEachGuest() [inlined] Run + 161 at callback.h:129 [opt]
    frame #2: 0x00000001033de96c Electron Framework`::ForEachGuest() + 76 at web_view_manager.cc:69 [opt]
    frame #3: 0x000000010122ef36 Electron Framework`::AreAnyGuestsCurrentlyAudible() + 166 at browser_plugin_embedder.cc:207 [opt]
    frame #4: 0x00000001017cf5ec Electron Framework`::OnAudioStateChanged() + 60 at web_contents_impl.cc:1587 [opt]
    frame #5: 0x000000010153751a Electron Framework`::UpdateStreams() + 426 at audio_stream_monitor.cc:195 [opt]
    frame #6: 0x0000000101537de3 Electron Framework`::UpdateStreamAudibleStateOnUIThread() + 323 at audio_stream_monitor.cc:157 [opt]
    frame #7: 0x00000001034a4a56 Electron Framework`::RunTask() [inlined] Run + 23 at callback.h:99 [opt]
    frame #8: 0x00000001034a4a3f Electron Framework`::RunTask() + 335 at task_annotator.cc:101 [opt]
    frame #9: 0x00000001034dd76f Electron Framework`::RunTask() + 639 at message_loop.cc:421 [opt]
    frame #10: 0x00000001034dddca Electron Framework`::DoWork() [inlined] DeferOrRunPendingTask + 474 at message_loop.cc:432 [opt]
    frame #11: 0x00000001034dddbe Electron Framework`::DoWork() + 462 at message_loop.cc:480 [opt]
    frame #12: 0x0000000103609cca Electron Framework`::RunWork() + 42 at message_pump_mac.mm:455 [opt]
    frame #13: 0x000000010360ec8a Electron Framework`base::mac::CallWithEHFrame(void () block_pointer) + 10 at call_with_eh_frame_asm.S:36

zcbenz · 2018-11-02T12:37:04Z

The crash turned out to be a mistake when refactoring.

All tests are passing now.

MarshallOfSound · 2018-11-02T12:54:41Z

@zcbenz The 4.0 backport failed, trop failed to report it because of a GitHub API error. Will work on a fix for trop for that but just a heads up for now 👍

Normally this means we send invalid annotations, so let's just send a failure without them. E.g. electron/electron#15541

deepak1556

LGTM 👍

deepak1556 · 2018-11-02T16:40:16Z

spec/fixtures/api/leak-exit-browserview.js

+app.on('ready', function () {
+  new BrowserView({})  // eslint-disable-line
+
+  process.nextTick(() => app.quit())


Any reason this has to be on the next tick ?

I'm trying to emulate the original crash case, having process.nextTick or not does not really affect the test.

#64) Normally this means we send invalid annotations, so let's just send a failure without them. E.g. electron/electron#15541

deepak1556 · 2018-11-03T08:07:21Z

@zcbenz just remembered, we had disabled a spec in 69 due to the same crash https://github.com/orgs/electron/projects/24#card-13687849, can you try enabling it in this PR ? Thanks!

zcbenz · 2018-11-08T12:24:47Z

@zcbenz just remembered, we had disabled a spec in 69 due to the same crash https://github.com/orgs/electron/projects/24#card-13687849, can you try enabling it in this PR ?

The test is timeout on my machine, after quitting the app directly I'm seeing the client_map_.empty() assertion. I'll look into it in another PR.

felixrieseberg

This makes sense to me and looks good. 👏 for the documentation in code, explains what's going on well!

release-clerk · 2018-11-08T15:57:32Z

Release Notes Persisted

Fix crash on exit when using BrowserView.

miniak · 2019-03-15T16:45:02Z

/trop run backport-to 3-1-x

trop · 2019-03-15T16:45:07Z

The backport process for this PR has been manually initiated,
sending your 1's and 0's to "3-1-x" here we go! :D

trop · 2019-03-15T16:45:24Z

I was unable to backport this PR to "3-1-x" cleanly;
you will need to perform this backport manually.

zcbenz added the target/4-0-x label Nov 2, 2018

zcbenz requested review from a team November 2, 2018 05:30

zcbenz force-pushed the destroy-webcontents-sync branch 2 times, most recently from 9f28853 to 7416c2e Compare November 2, 2018 08:34

zcbenz force-pushed the destroy-webcontents-sync branch 3 times, most recently from d5b5a65 to 2ed1a11 Compare November 2, 2018 11:03

MarshallOfSound added target/4-0-x and removed target/4-0-x labels Nov 2, 2018

MarshallOfSound added a commit to electron/trop that referenced this pull request Nov 2, 2018

fix: try mark as failure without annotations when the GitHub API fails

0dd9f65

Normally this means we send invalid annotations, so let's just send a failure without them. E.g. electron/electron#15541

MarshallOfSound mentioned this pull request Nov 2, 2018

fix: try mark as failure without annotations when the GitHub API fails electron/trop#64

Merged

deepak1556 approved these changes Nov 2, 2018

View reviewed changes

fix: destroy WebContents synchronously on shutdown

d6c0b35

zcbenz force-pushed the destroy-webcontents-sync branch from 2ed1a11 to d6c0b35 Compare November 8, 2018 12:25

zcbenz removed the target/4-0-x label Nov 8, 2018

zcbenz mentioned this pull request Nov 8, 2018

fix: destroy WebContents synchronously on shutdown (4-0-x) #15640

Merged

codebytere approved these changes Nov 8, 2018

View reviewed changes

felixrieseberg approved these changes Nov 8, 2018

View reviewed changes

codebytere merged commit 746beb0 into master Nov 8, 2018

codebytere deleted the destroy-webcontents-sync branch November 8, 2018 15:57

deepak1556 pushed a commit that referenced this pull request Nov 26, 2018

fix: destroy WebContents synchronously on shutdown (#15541)

1001f26

deepak1556 mentioned this pull request Nov 26, 2018

fix: switch to mojo proxy resolver (3-1-x) #15813

Merged

5 tasks

deepak1556 pushed a commit that referenced this pull request Nov 26, 2018

fix: destroy WebContents synchronously on shutdown (#15541)

9a9b382

miniak mentioned this pull request Mar 15, 2019

[wip] fix: destroy WebContents synchronously on shutdown (3-1-x) #17395

Closed

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

fix: destroy WebContents synchronously on shutdown #15541

fix: destroy WebContents synchronously on shutdown #15541

zcbenz commented Nov 2, 2018

zcbenz commented Nov 2, 2018

zcbenz commented Nov 2, 2018

MarshallOfSound commented Nov 2, 2018

deepak1556 left a comment

deepak1556 Nov 2, 2018

zcbenz Nov 2, 2018

deepak1556 commented Nov 3, 2018

zcbenz commented Nov 8, 2018

felixrieseberg left a comment

release-clerk bot commented Nov 8, 2018

miniak commented Mar 15, 2019

trop bot commented Mar 15, 2019

trop bot commented Mar 15, 2019

fix: destroy WebContents synchronously on shutdown #15541

fix: destroy WebContents synchronously on shutdown #15541

Conversation

zcbenz commented Nov 2, 2018

Description of Change

Checklist

Release Notes

zcbenz commented Nov 2, 2018

zcbenz commented Nov 2, 2018

MarshallOfSound commented Nov 2, 2018

deepak1556 left a comment

Choose a reason for hiding this comment

deepak1556 Nov 2, 2018

Choose a reason for hiding this comment

zcbenz Nov 2, 2018

Choose a reason for hiding this comment

deepak1556 commented Nov 3, 2018

zcbenz commented Nov 8, 2018

felixrieseberg left a comment

Choose a reason for hiding this comment

release-clerk bot commented Nov 8, 2018

miniak commented Mar 15, 2019

trop bot commented Mar 15, 2019

trop bot commented Mar 15, 2019