209568 - [FIX]Freetype code leaks all over

Status: RESOLVED FIXED

(Core Graveyard :: GFX: Gtk, defect, P1)

Description

[Boris Zbarsky [:bzbarsky]]

I haven't done a particularly thorough examination; it just failed to compile
due to some leak-proofing patches in my tree... Short list:

1)  nsFreeTypeGetFaceID leaks if Init() fails on a face
2)  LoadNodeTable and GetFontNames both leak the arrayFC array (the getter
    addrefs, they do not releas).
3)  The same functions, if the array leak is fixed, leak every single entry in
    the array, since ElementAt on an nsISupportsArray returns an _addrefed_
    pointer.

I may take a shot at converting some of this code to nsCOMArray at some point,
but in the meantime, we could try fixing the leaks by the simple expedient of
actually using nsCOMPtrs for owning references and using QueryElementAt on the
nsISupportsArray....

Comment 1

[Boris Zbarsky [:bzbarsky]]

Attachment: Some possible leak fixes

We probably want to read this code over and fix any other leaks we find...

Comment 2

[Boris Zbarsky [:bzbarsky]]

Attachment: Better patch

Actually, looks like the faces never got leaked.  The arrays themselves did,
though.  Fixing the leaks as in attachment 125759 [details] [diff] [review] brought up the issue of
refcounted objects being just deleted and being passed around with a zero
refcount (once the refcount from the leaked array was removed).

This patch just makes everything use refcounting to ensure that life is happy.

Comment 3

[Boris Zbarsky [:bzbarsky]]

Comment on attachment 125792 [details] [diff] [review]
Better patch

Going for the double-Brian review.  ;)

Comment 4

[Boris Zbarsky [:bzbarsky]]

I guess I should take this....

Comment 5

[kill this account]

Comment on attachment 125792 [details] [diff] [review]
Better patch

The logic is unchanged and looks okay to me so r=[e-mail]

Comment 6

[Brian Ryner (not reading)]

Comment on attachment 125792 [details] [diff] [review]
Better patch

sr=me

Comment 7

[Boris Zbarsky [:bzbarsky]]

Checked in.