473548 - leak 2 nsDisplayBackground when running each 411585-*.html reftest in bugs/reftest.list

Status: VERIFIED FIXED

(Core :: Printing: Output, defect, P3)

Description

[Serge Gautherie (:sgautherie)]

[Mozilla/5.0 (Windows; U; Windows NT 5.0; en-US; rv:1.9.2a1pre) Gecko/20090114 Minefield/3.2a1pre] (home, optim default) (W2Ksp4)
(http://hg.mozilla.org/mozilla-central/rev/e0ed5852481a)

After fixing the nsMathMLContainerFrame leak of bug 458844,
its (6) nsDisplayBackground leak remains.

Run with only
{
# bugs/
include bugs/reftest.list
}

Then I tried to narrow that list down:
{
Whole file: leaks 6

start to 369882.xul: leaks 0
370353-1.html to end: leaks 3

370353-1.html to 405584-1.html: leaks 0
405952-1.html to end: leaks 0
}
There are more than 1,000 tests and I don't know how to "easily" find out which ones need to be run to trigger the leak :-/

Comment 1

[Serge Gautherie (:sgautherie)]

(In reply to comment #0)
> 370353-1.html to end: leaks 3

I don't know what happened: this one is still 6.

> 405952-1.html to end: leaks 0

I had not saved the list file: this one is still 6.

The culprits are
{
== 411585-1.html 411585-1-ref.html
== 411585-2.html 411585-2-ref.html
fails == 411585-3.html 411585-3-ref.html # bug 426909
}
which leak 2 each.
(Maybe 1 from the test and 1 from the ref !?)

Comment 2

[Joel Maher ( :jmaher ) (UTC -8)]

As a note, I have ran into issues while testing on fennec.  This bugs directory is so large I have had to split it up into 10 .list files and run each one standalone.  There isn't much I do other than some head & tail stuff to create smaller .list files, but if I can help out in any way, let me know.

Comment 3

[Serge Gautherie (:sgautherie)]

(In reply to comment #2)
> I have had to split it up into 10 .list files

That is unrelated to this bug, but I guess the only solution would be to pre-process the list files to do that at build time.

Comment 4

[Serge Gautherie (:sgautherie)]

(In reply to comment #2)
> and run each one standalone.

And build a list of all the lists, in order to run them individually instead of by includes as it is now.

Comment 5

[Joel Maher ( :jmaher ) (UTC -8)]

as a reference for the fennec stuff I have entered a bug 471676 last year to find a method for running the tests in smaller chunks.

Comment 6

[David Baron :dbaron: (⌚️UTC-5, no longer working on Mozilla)]

Attachment: stack for constructor call

I'm seeing two constructor calls at the same address with no matching destructor.  Here's the part of the stack in common between those two calls.

Comment 7

[David Baron :dbaron: (⌚️UTC-5, no longer working on Mozilla)]

Attachment: patch

Comment 8

[David Baron :dbaron: (⌚️UTC-5, no longer working on Mozilla)]

Fixed: http://hg.mozilla.org/mozilla-central/rev/0eab2ed55fac

Comment 9

[David Baron :dbaron: (⌚️UTC-5, no longer working on Mozilla)]

Comment on attachment 356970 [details] [diff] [review]
patch

Trivial fix to avoid skipping virtual destructors so that we fix the leak statistics.

Comment 10

[Serge Gautherie (:sgautherie)]

[Mozilla/5.0 (Windows; U; Windows NT 5.0; en-US; rv:1.9.2a1pre) Gecko/20090117 Minefield/3.2a1pre] (home, optim default) (W2Ksp4)
(http://hg.mozilla.org/mozilla-central/rev/184eadf4e185 + bug 446300 patch)

V.Fixed

Comment 11

[David Baron :dbaron: (⌚️UTC-5, no longer working on Mozilla)]

http://hg.mozilla.org/releases/mozilla-1.9.1/rev/1947fde473c5