970042 - Use fallible allocation in the Windows widget code

Status: RESOLVED FIXED

(Core :: Widget: Win32, defect)

Comment 1

[(no longer active)]

Attachment: Use fallible allocation in the Windows widget code

Comment 2

[Robert O'Callahan (:roc) (email my personal email if necessary)]

Comment on attachment 8373010 [details] [diff] [review]
Use fallible allocation in the Windows widget code

Review of attachment 8373010 [details] [diff] [review]:
-----------------------------------------------------------------

Why are we doing this? Do you have some reason to believe these are under Web content control?

Comment 3

[(no longer active)]

These arrays are used to compute the regions that we want to paint, which is presumably semi-controllable from content based on the invalidations that it causes.  But as I was typing this I realized that the size of the screen is obviously not controlled by content so there is an upper bound on the number of regions the content can invalidate, and I can't convince myself properly that content can create enough invalidations to cause an OOM here (unless we're already close to OOMing anyways.)  So I guess a new patch is in order.

Comment 4

[(no longer active)]

Attachment: Stop checking the result of nsTArray::SetLength in the Windows widget code; r=roc

Comment 5

[Robert O'Callahan (:roc) (email my personal email if necessary)]

Comment on attachment 8373069 [details] [diff] [review]
Stop checking the result of nsTArray::SetLength in the Windows widget code; r=roc

Review of attachment 8373069 [details] [diff] [review]:
-----------------------------------------------------------------

We do region simplification for performance reasons which should prevent content from making these regions arbitrarily memory-hungry.

Comment 6

[(no longer active)]

https://hg.mozilla.org/integration/mozilla-inbound/rev/1827e672f179

Comment 7

[Carsten Book [:Tomcat]]

https://hg.mozilla.org/mozilla-central/rev/1827e672f179