FireFox V53.0 just grinds to a halt




2 years ago
2 years ago


(Reporter: david, Unassigned)


53 Branch

Firefox Tracking Flags

(Not tracked)



(1 attachment)



2 years ago
User Agent: Mozilla/5.0 (Windows NT 10.0; rv:53.0) Gecko/20100101 Firefox/53.0
Build ID: 20170413192749

Steps to reproduce:

Run FireFox (I am using V53.0 32-bit on Windows 10) with several windows and tabs open from previous session.

Actual results:

CPU usage by FireFox is at about 52% and memory used by the program is at ~650MB (out of 4GB) so it is neither CPU-, nor memory-bound
Works fine for a while, and then starts to take longer and longer to respond, until the point where it simply stops.  This can take anything from a few minutes to an hour to cease responding.  If I catch it soon enough, then using File-->Exit and restarting gets me back to a working position - for a while - and then the same slowing down and grinding to a halt occurs.  If I am not on the ball, then the only way to proceed is to kill the process from the Task Manager and restart.
Watching YouTube videos or downloading large files _appear_ to make the slowing down happen sooner although this could be a misconception.
This performance issue arose initially with V51.0.

Expected results:

FireFox should remain active and responsive even if left open for extended periods of time.  Earlier versions have remained active for several days without any deterioration in response.
Hi David,

Thanks for reporting this issue. Can you send us a Performance Profile of the issue? Instructions on how to do that can be followed here:
Component: Untriaged → General
Flags: needinfo?(david)

Comment 2

2 years ago
Created attachment 8863936 [details]
Gecko Profiler log (compressed)
Flags: needinfo?(david)

Comment 3

2 years ago
I installed Nightly "55.0a1 (2017-05-02) (32-bit)" and have run it using the same set of windows and tabs which I have been using previously with FireFox.  Having run Nightly for about an hour, playing several YouTube videos and loading a number of other web pages, I have not noticed any slowing down in the performance of the software.  I have attached the log from the Performance Profile, although I am not sure that this will show anything useful at this time.  I will continue and leave the program running and rerun more tests tomorrow.
I noticed that this version of Nightly is using significantly less CPU than I have been using with FireFox V53.0, V52.0 and V51.0 which invariably ran at somewhere around 56% CPU.  Nightly is running consistently with between 6% and 15% CPU - again, for reference, using the same set of windows and tabs as I have been using with FireFox.
Could something that has been causing the performance issue perhaps already have been addressed?

Comment 4

2 years ago
I left it running overnight and it was still working this morning.  I have updated to Nightly 55.0a1 (2017-05-03) (32-bit) and have run as many YouTube videos as I have had the time to do, with the result that there has been no noticeable degradation in performance.

I did notice in Task Manager that there are several background Nightly tasks running (five to be precise) as well as the main one in the foreground.  I have eight Nightly windows open.  A snapshot of the percentage CPU usage of each task is:
Foreground task: 11.9%
Background tasks: 11.6%, 4.4%, 0.8% 0.3% and 12.1% (I realise that this information is probably of no use whatsoever, but as the main program has not displayed the performance issue that FireFox has been demonstrating, I wanted to provide something at least.)

Comment 5

2 years ago
I went away for a few days leaving Nightly 55.0a1 (2017-05-03) (32-bit) running and it was still fine upon my return yesterday.
Today, I have been running FireFox 53.0 (32-bit) and it has for the most part been running OK.  The only problem I have experienced is that it lost its plot with the graphics - tabs all went blank so I had to restart.  I had a dialog up to upgrade to 53.0.2 I think it was, so I will update to that in a short while.
You need to log in before you can comment on or make changes to this bug.