296058 - SVG+JS allows DoS against the UI event handling

Status: RESOLVED WORKSFORME

(Core :: SVG, defect)

Description

[Henri Sivonen (:hsivonen)]

Build Id: Deer Park 20050530 (on OS X 10.3.9 on 533 MHz G4)

Steps to reproduce:
1) Load http://www.hixie.ch/tests/adhoc/svg/perf/001.xml

Actual results:
The image animates. The mouse cursor turns into the Spinning Beach Ball of
Death. The UI does not respond. Can't leave the page.

Expected results:
Expected the UI to remain responsive.

Comment 1

[Boris Zbarsky [:bzbarsky]]

This may be platform-specific, but on Linux almost all the time on this testcase
seems to be inside cairo...

Comment 2

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

DO we really want to try to address this sort of thing?

Comment 3

[Boris Zbarsky [:bzbarsky]]

There are two separate questions there:

1)  Do we want to address performance bugs?  Imo, yes.
2)  Do we want to address cases when we go away from the even loop forever by
doing things in smaller chunks if possible?  Imo again yes, but this is harder.

Comment 4

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

This doesn't hang for me on trunk. I get about 3.5 frames per second, which
isn't great, but isn't a hang.

Comment 5

[Boris Zbarsky [:bzbarsky]]

The hang is almost certainly Mac-specific (I don't hang on Linux either, but the
Mac event queues have slightly different event priorities than the Linux and
Windows ones, if I recall correctly).

Comment 6

[Simon Fraser [no longer active]]

Can we get a stack for the hang? Run 'sampler' on the process.

Comment 7

[Jonathan Watt [:jwatt]]

Brian, can you check this testcase doesn't hang firefox (the circle animates) on Mac? Make sure you don't mind killing firefox before trying it of course!

Comment 8

[Brian Ewins]

(In reply to comment #7)
> Brian, can you check this testcase doesn't hang firefox (the circle animates)
> on Mac? Make sure you don't mind killing firefox before trying it of course!
> 

No hang on either trunk or branch
Mozilla/5.0 (Macintosh; U; Intel Mac OS X; en-GB; rv:1.8.1.10) Gecko/20071115 Firefox/2.0.0.10
Mozilla/5.0 (Macintosh; U; Intel Mac OS X 10.5; en-US; rv:1.9b1) Gecko/2007110903 Firefox/3.0b1

About 4.6 fps on trunk and 4.4 fps on branch. There is a rendering problem on trunk - the text isn't clipped correctly, I have to switch windows to force a redraw to see the timer. Different problem though, maybe bug 321470.

Comment 9

[Jonathan Watt [:jwatt]]

Can you check the text here and bug 321470 again now that cairo has been updated Brian.

Comment 10

[Brian Ewins]

There's still no hang in a build I made last night 
Mozilla/5.0 (Macintosh; U; Intel Mac OS X 10.5; en-US; rv:1.9b2pre) Gecko/2007113023 Minefield/3.0b2pre

However the text is still busted, and it does seem to be related to bug 321470. I noticed by accident that the bustage didn't happen when I used a different default font (ie this bug is a WFM, the reported problem has gone). The default font with the problem was Times. I'll put more info about the font issue on 321470.