397793 - Memory leak of documents and docshells (for Sunbird)

Status: VERIFIED DUPLICATE of bug 426966

(Calendar :: Sunbird Only, defect)

Description

[Gary Kwong [:gkw] [:nth10sd] (NOT official MoCo now)]

User-Agent:       Mozilla/5.0 (Macintosh; U; Intel Mac OS X; en-US; rv:1.8.1.7) Gecko/20070914 Firefox/2.0.0.7
Build Identifier: Mozilla/5.0 (Macintosh; U; Intel Mac OS X; en-US; rv:1.8.1.8pre) Gecko/20070926 Calendar/0.7pre

Memory leak found when testing Sunbird.

http://www.squarefree.com/2007/09/20/firefox-memory-usage-and-memory-leak-news/

Reproducible: Always

Steps to Reproduce:
1. run ./sunbird (or sunbird-bin) && perl /Applications/leak-gauge.pl nspr_leaks.log
2. close sunbird immediately after it opens.
Actual Results:  
*** Calendar schema version is: 7
Leaked outer window 19a9800 at address 19a9800.
Leaked outer window 1dd0ce0 at address 1dd0ce0.
Leaked inner window 1dd40a0 (outer 1dd0ce0) at address 1dd40a0.
 ... with URI "about:blank".
Leaked inner window 1d4e9e0 (outer 19a9800) at address 1d4e9e0.
 ... with URI "chrome://calendar/content/hiddenWindow.xul".
Summary:
Leaked 4 out of 4 DOM Windows
Leaked 0 out of 30 documents
Leaked 0 out of 2 docshells

Expected Results:  
no leaks.

I placed leak-gauge.pl in "/Applications/".

Comment 1

[Gary Kwong [:gkw] [:nth10sd] (NOT official MoCo now)]

Just noticed that David Baron had mentioned that "The current logging is targeted more at browsing than email: it may show some leaks in Thunderbird as well, but it may or may not be a good sample of leaks in Thunderbird."

http://dbaron.org/log/2006-01#e20060110a

Not sure if the script is meant to work for Sunbird as well. But such a script could be a good way of looking for memory leaks in Sunbird IMHO.

Comment 2

[Gary Kwong [:gkw] [:nth10sd] (NOT official MoCo now)]

Mozilla/5.0 (Macintosh; U; Intel Mac OS X 10.5; en-US; rv:1.9b3pre) Gecko/2008010617 Calendar/0.6a1

Tested and confirming on a debug Calendar app built from cvs.

Will be attaching output in gdb.

Comment 3

[Gary Kwong [:gkw] [:nth10sd] (NOT official MoCo now)]

Attachment: gdb output

Upon opening then closing Sunbird, this gdb output shows the leak objects.

(Leaked 1,044 from mFreeCount and 2 from mAdoptFreeCount)

Comment 4

[Gary Kwong [:gkw] [:nth10sd] (NOT official MoCo now)]

(not nominating for 0.8 because the debug build in comment #2 is a trunk one, and not from the branch)

Comment 5

[Michiel van Leeuwen (email: mvl+moz@)]

Those mFreeCount and mAdoptFreeCount are only start from internal string code, not things that are leaks. And it doesn't surprise me that strings are leaked when entire windows are leaked...
Did your test include the patch for bug 410168?

Comment 6

[Gary Kwong [:gkw] [:nth10sd] (NOT official MoCo now)]

I compiled the cvs build today, date stamp is 2008010617 (6th Jan)

The patch for bug 410168 was checked in on 3rd Jan, around three days ago, so yes the test should have included the patch.

(Apologies for the misintepretation, I'm still new to debug builds, not sure what to look out for.)

Comment 7

[Gary Kwong [:gkw] [:nth10sd] (NOT official MoCo now)]

nominating blocking on par with "no leaks" release goal.

any relation to bug 413296?

Comment 8

[Simon Paquet [:sipaq]]

We're not going to block 0.8 for this bug, but we'll of course accept patches.

Comment 9

[Gary Kwong [:gkw] [:nth10sd] (NOT official MoCo now)]

Duping in favour of bug 426966.