626177 - flash ends up in lower quality sometimes

Status: RESOLVED FIXED

(Core Graveyard :: Plug-ins, defect)

Description

[Vladimir Vukicevic [:vlad] [:vladv] (needinfo me, slow to respond)]

I've noticed this on a few sites -- the rdio collections circle/bubble graph, and now the flash at http://www.logitech.com/en-us/speakers-audio -- the text is not antialiased, though it looks like it tried to do AA but was instead forced to 100% opacity somehow.  Could be flash thinking it has a surface with an alpha channel and us not giving it one, not sure.

Bad (Firefox 4 nightly): http://i.imgur.com/VPLCs.png
Good: http://i.imgur.com/5bHKJ.png

Comment 1

[Alice0775 White]

Regression window:
Works;
http://hg.mozilla.org/mozilla-central/rev/212a391d3b79
Mozilla/5.0 (Windows NT 6.1; WOW64; rv:2.0b8pre) Gecko/20101109 Firefox/4.0b8pre ID:20101110052514
Fails:
http://hg.mozilla.org/mozilla-central/rev/bdbef533364f
Mozilla/5.0 (Windows NT 6.1; WOW64; rv:2.0b8pre) Gecko/20101109 Firefox/4.0b8pre ID:20101110081923
Pushlog:
http://hg.mozilla.org/mozilla-central/pushloghtml?fromchange=212a391d3b79&tochange=bdbef533364f

Comment 2

[Benjamin Smedberg]

Did you try this after bug 611698 landed (past two days)?

Comment 3

[Alice0775 White]

(In reply to comment #2)
> Did you try this after bug 611698 landed (past two days)?
Yes, Latest m^c nightly.
And confirmed again on http://hg.mozilla.org/mozilla-central/rev/413eb50d4b81
Mozilla/5.0 (Windows NT 6.1; WOW64; rv:2.0b10pre) Gecko/20110116 Firefox/4.0b10pre ID:20110116030325

Comment 4

[Ryan VanderMeulen [:RyanVM]]

This sounds like the espn.com issue I reported in bug 611698.

Comment 5

[Vladimir Vukicevic [:vlad] [:vladv] (needinfo me, slow to respond)]

For me, turning Cleartype off doesn't affect the bad rendering -- still stays bad.  Problem sounds/looks to be very similar to bug 611698, though.

Comment 6

[Vladimir Vukicevic [:vlad] [:vladv] (needinfo me, slow to respond)]

Also, the espn.com poll looks fine to me here.  Marking this related to 611698, though.

Comment 7

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

If you try with OOPP disabled and Cleartype manually disabled, do you see the same problem?

Comment 8

[Chris B.]

Confirmed - seeing issue mentioned in comment 0.
Although, for me, all sites mentioned in bug 611698 work fine.

Comment 9

[Alice0775 White]

Attachment: reduced html

FYI
Switching tab or scrolling up/down helps for upper sample(espn.com).

Comment 10

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

To answer my own question: no, that does not trigger the problem.

Comment 11

[Alice0775 White]

(In reply to comment #7)
> If you try with OOPP disabled and Cleartype manually disabled, do you see the
> same problem?

OOPP disabled : The problem does not happen
Cleartype manually disabled : The problem happens.

Comment 12

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

I meant both at the same time.

Comment 13

[Alice0775 White]

(In reply to comment #12)
> I meant both at the same time.

Disabled both;
http://www.logitech.com/en-us/speakers-audio : The problem does not happen
espn.com: The problem does not happen, but the font is thin than that of disabled OOPP only

Comment 14

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

OK I've figured out what the problem is. The problem is that Flash recomposites its entire internal buffer over our own irrespective of what rectangle we asked it to paint. In PluginInstanceChild::PaintRectToSurface we only clear the rectangle that we're asking Flash to paint into. So Flash ends up recompositing text outside the repaint region over and over itself, which eventually saturates any pixels that were partially touched by the text.

The best way to fix this is to set the GDI clip rect to the rectangle we expect Flash to paint into.

Comment 15

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

Attachment: fix

This fixes it.

Comment 16

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

Setting the clip rect like this might speed up Flash's final BitBlt too, for some performance gain.

Comment 17

[Ryan VanderMeulen [:RyanVM]]

Patch applied locally looks great, Rob.

Comment 18

[Matt Woodrow (:mattwoodrow)]

http://hg.mozilla.org/mozilla-central/rev/aa611c72c337