271201 - Variable size memory leak when new window or tab is opened and closed

Status: RESOLVED DUPLICATE of bug 258567

(Firefox :: General, defect)

Description

[matt235]

User-Agent:       Mozilla/5.0 (Windows; U; Windows NT 5.0; en-US; rv:1.7.5) Gecko/20041107 Firefox/1.0
Build Identifier: Mozilla/5.0 (Windows; U; Windows NT 5.0; en-US; rv:1.7.5) Gecko/20041107 Firefox/1.0

Whenever a new window or tab of Firefox is opened, to any webpage, more memory
is taken up.  Minimizing does not help, nor does quitting Firefox entirely.  The
amount of memory leaked varies from roughly 5k to 100k per open and close.  I am
running Windows 2000 with all available updates, and I know someone who has the
same problem on a similarly patched Windows XP.  The only extensions I have
installed are Googlebar 0.9.0.30 and DOM Inspector 1.0.  The leak is large
enough to make the system sluggish within a half-day of normal use.

Reproducible: Always
Steps to Reproduce:
1.Open a new window or tab of Firefox
2.Close the window or tab
3.

Actual Results:  
More memory was reported in use by Windows than before.

Expected Results:  
No more memory should have been left in use.

Comment 1

[Peter da Silva]

Seems to happen on Mac OS X, as reported by

while 1
ps waux | grep Fire | grep -v grep
sleep 1
end

Comment 2

[Steve Chapel]

Dupe of bug 131456?

Comment 3

[matt235]

(In reply to comment #2)
> Dupe of bug 131456?

No, this is a very different bug.  Bug 131456 regards memory not being freed
until the browser is closed; this bug is much more serious in the sense that to
free the memory you must restart your computer entirely.

I *really* wish someone on the dev team would at least confirm this bug.  I'm
part of the computer management team for Stanford and this is one of the biggest
things preventing me from recommending near-exclusive use of Firefox on campus
(that's 15,000 people).

Comment 4

[Steve Chapel]

When you say "More memory was reported in use by Windows than before," how
exactly did you determine that? Please give the raw data, not an interpretation
of the data.

Comment 5

[matt235]

(In reply to comment #4)
> how exactly did you determine that?

I determined this by leaving Windows Task Manager open.  In Windows 2000 it
reports the memory usage in kB, not MB as in XP.  My apologies for the omission.

Comment 6

[Steve Chapel]

I'm not sure which number to watch. Is it in the Mem Usage column? If so, which
process should I be watching when I try to reproduce this bug?

Also, could you try reproducing this bug in a recent nightly of Firefox so we
don't spend lots of time tracking down a bug that's already fixed? Thanks!

Comment 7

[matt235]

(In reply to comment #6)
> I'm not sure which number to watch. Is it in the Mem Usage column? If so, which
> process should I be watching when I try to reproduce this bug?

Watch the total Mem Usage, at the bottom of the task manager.  It won't show up
in the column because the process using the memory has disappeared.

I'll double-check with a nightly as soon as I get time.

Comment 8

[Steve Chapel]

I'm using Windows XP, and I don't see any total Mem Usage in the Windows Task
Manager. I tried looking at the commit charge instead. I can't see any increase
after I open lots of tabs or windows then quit Mozilla trunk build 2005021105.

Comment 9

[matt235]

(In reply to comment #8)
> I'm using Windows XP, and I don't see any total Mem Usage in the Windows Task
> Manager. I tried looking at the commit charge instead. I can't see any increase
> after I open lots of tabs or windows then quit Mozilla trunk build 2005021105.

In XP that's what it calls it.  As I said, I haven't tried the new trunk build
yet, but I'm guessing the reason you don't see the memory usage difference is
because (as mentioned previously) XP displays the commit charge in MB, not kB. 
On my machine each open-close appears to orphan between 5 and 25 kB of memory,
so it would take a fair number of open-closes to see it in XP.

Comment 10

[Steve Chapel]

(In reply to comment #9)
> In XP that's what it calls it.  As I said, I haven't tried the new trunk build
> yet, but I'm guessing the reason you don't see the memory usage difference is
> because (as mentioned previously) XP displays the commit charge in MB, not kB. 
> On my machine each open-close appears to orphan between 5 and 25 kB of memory,
> so it would take a fair number of open-closes to see it in XP.

I finally see what number you're referring to! In the status bar of the Windows
Task Manager, yes, Windows XP displays the commit charge in MB. I was looking at
the commit charge in the Performance tab, where it is listed in KB. I actually
saw that number *decrease* when I followed your steps to reproduce.

Comment 11

[Robin Monks]

Dupe of bug 258567.  Reporter, if you feel this was done in error please REOPEN.

*** This bug has been marked as a duplicate of 258567 ***