Closed Bug 1274725 Opened 8 years ago Closed 8 years ago

Remaining "Nightly Web Content" process after browser have been closed

Categories

(Core :: DOM: Content Processes, defect)

Unspecified
macOS
defect
Not set
normal

Tracking

()

RESOLVED DUPLICATE of bug 1268559
Tracking Status
e10s ? ---

People

(Reporter: clement.lefevre, Unassigned)

References

Details

(Keywords: nightly-community, Whiteboard: [mozfr-community] btpp-followup-2016-06-06)

Attachments

(1 file)

I just noticed on my computer running OS X that there was a remaining process from Nightly called "Nightly Web Content" long time after I closed the browser. Its parent process was not alive anymore, and the system was telling its ppid is 1 (launchd on OS X).

The process was the following one:

mbp :: ~ » ps aux | grep 20045
clement         20045   5,8  8,5  5507868 888756   ??  S     4:19    31:05.97 /Applications/FirefoxNightly.app/Contents/MacOS/plugin-container.app/Contents/MacOS/plugin-container -greomni /Applications/FirefoxNightly.app/Contents/Resources/omni.ja -appomni /Applications/FirefoxNightly.app/Contents/Resources/browser/omni.ja -appdir /Applications/FirefoxNightly.app/Contents/Resources/browser 20042 gecko-crash-server-pipe.20042 org.mozilla.machname.1582198594 tab

I did gathered a process dump before killing it. You can find it attached.

As additional informations, the process was not declared as zombie, was still consuming CPU and memory usage was increasing over the time and would have indefinitely if I didn't killed the process.
Whiteboard: [mozfr-community][nightly-community]
OS: Unspecified → Mac OS X
Mike, wdyt?
Flags: needinfo?(mconley)
Whiteboard: [mozfr-community][nightly-community] → [mozfr-community][nightly-community] btpp-followup-2016-06-06
Sounds similar to something dolske reported in bug 1276330. Adding this to the e10s triage.
tracking-e10s: --- → ?
Flags: needinfo?(mconley)
See Also: → 1276330
That process sample is interesting. If I'm interpreting it right, it looks like some page is doing a (sync?) XHR on the pagehide event that's being fired on TabChild teardown.
Blocks: 1268559
No longer blocks: 1268559
Depends on: 1268559
Status: NEW → RESOLVED
Closed: 8 years ago
No longer depends on: 1268559
Resolution: --- → DUPLICATE
I managed to reproduce something like this. I made a testcase that does a sync XHR to a slow web site in a pagehide event handler. Then I also opened Treeherder along with this tab. While Treeherder is still loading, I hit Ctrl-Q. The parent process crashes in a shutdown hang while the child lives forever. I'm not sure why I get a shutdown hang on Linux while Mac people don't seem to get that.

Anyway, my shutdown hang is in the compositor shutdown code, so I'm not quite confident this is a dupe.
Whiteboard: [mozfr-community][nightly-community] btpp-followup-2016-06-06 → [mozfr-community] btpp-followup-2016-06-06
You need to log in before you can comment on or make changes to this bug.

Attachment

General

Created:
Updated:
Size: