gal: do we rely on surfaces to be clean? [7:36pm] gal: because I bet you gralloc doesn't guarantee that [7:36pm] gal: and I didn't see a memset there [7:37pm] mattwoodrow: OSX shmem was guaranteeing that [7:37pm] mattwoodrow: and we had failures if I got rid of it [7:37pm] gal: ok, most the time there won't be alpha [7:37pm] gal: so this is good enough for me [7:38pm] mattwoodrow: Also, we clear the surface in ContentClientIncremental::BeginPaintBuffer [7:38pm] mattwoodrow: so i'm not why the memset would be needed at all [7:38pm] mattwoodrow: debugging fail at the time maybe [7:38pm] bbondy left the chat room. (Quit: Going offline, see ya! (www.adiirc.com)) [7:39pm] gal: lets do 2 patches, one to yank the memset [7:39pm] gal: another to reuse the buffer [7:39pm] gal: we can try them separately [7:39pm] gal: or I can do if (aContent != gfaSurface::CONTENT_COLOR) { [7:39pm] gal: dealers choice [7:40pm] mattwoodrow: try yanking it entirely

Now passes all tests.

https://hg.mozilla.org/integration/mozilla-inbound/rev/c32dc74cd74a https://hg.mozilla.org/integration/mozilla-inbound/rev/ef78caaf7301

https://hg.mozilla.org/mozilla-central/rev/c32dc74cd74a https://hg.mozilla.org/mozilla-central/rev/ef78caaf7301

Nice work with the mac fix.