227963 - [gtk2] native window code leaks

Status: RESOLVED FIXED

(Core Graveyard :: Plug-ins, defect)

Description

[Christopher Blizzard (:blizzard)]

There's a leak in the native gtk2 window code.  This causes a lot of problems
during destruction.  Will attach a patch in a moment.

Comment 1

[Christopher Blizzard (:blizzard)]

Attachment: patch

Should be using an nsCOMPtr<> as well as using getter_AddRefs().

Comment 2

[Robin Lu]

seems like the leak comes from the java plugin checking code. Add related people
to the cc list.

Comment 3

[Jack Lu]

Robin,

> seems like the leak comes from the java plugin checking code
Could you elaborate your speculation?

Comment 4

[Robin Lu]

Comment on attachment 137146 [details] [diff] [review]
patch

It is a little strange that we have to handle the plug_removed. I have checked
the code of gtksocket, seems like they need to provide a default value as true
to the result before emit the signal plug_remove.  It could be a bug of
gtksocket. Anyway, this patch is OK to me.

Comment 5

[Robin Lu]

Xiaobin,

You can have a look of the patch attached by Blizzard. 

Comment 6

[Jack Lu]

Wouldn't it be nicer if AddRef called from nsIPluginInstance::GetPeer function?

Comment 7

[Christopher Blizzard (:blizzard)]

I talked to own about the default for plug_removed because I thought it was odd
as well.  He says that this was the behaviour of gtk 1.2, so it's preserved in
the gtk2 code to maintain compatibility.

Comment 8

[Christopher Blizzard (:blizzard)]

Comment on attachment 137146 [details] [diff] [review]
patch

Good solid leak fix for 1.6 to let people write gtk2 plugins.

Comment 9

[Asa Dotzler [:asa]]

Comment on attachment 137146 [details] [diff] [review]
patch

a=asa (on behalf of drivers) for checkin to Mozilla 1.6

Comment 10

[Christopher Blizzard (:blizzard)]

Checked in before 1.6.  Thanks!

Comment 11

[Christopher Blizzard (:blizzard)]

Fixed.