[OpenLayers-Dev] Firefox crashes

Zac Barton zacbarton at gmail.com
Mon Apr 16 19:40:36 EDT 2007


Hi Chris

I found bad SVG performance on FF while working a project some time
ago that used very large values.

I think this is why;
http://developer.mozilla.org/en/docs/SVG_in_Firefox_1.5

Coordinate range
If your content has geometry with a large coordinate range, you need
to watch out for problems caused by cairo's internal use of a 16.16
bit fixed point representation for calculations. Cairo doesn't clip
primitives before rasterization, so final coordinates after
transformation that exceed the range -32678 to 32677 will cause
rendering errors and possibly very slow performance.


I managed to workaround my issue by translating the values relative to
0,0; transform="translate(-532603.60,172348.90)".


Zac

On 4/14/07, Christopher Schmidt <crschmidt at metacarta.com> wrote:
> To those who have had firefox crashes:
>
> I think I was just able to work out the core reason for the Firefox
> crashes we were seeing. Firefox's SVG implementation seems to strongly
> dislike very large values. I've just made a commit that changes the
> renderer to map all values into local pixel space.
>
> This should fix the Firefox crashes. With WFS, Tim Schaub reported that
> the patch on http://trac.openlayers.org/ticket/663 helped as well.
>
> If you have been able to crash / lock up firefox in the past, please try
> again with trunk. If you're still getting crashes, please apply
> http://trac.openlayers.org/attachment/ticket/663/wfs-hangs-firefox.patch
> and then let the list know that it was needed.
>
> I'd especially to get Stephen Woodbridge to report back on this: I can
> no longer reproduce his problem on
> http://dev.openlayers.org/sandbox/crschmidt/541/examples/ld-testv.html ,
> which seems like a good sign, but other users are always better bug
> testers than me :)
>
> Thanks,
> --
> Christopher Schmidt
> MetaCarta
> _______________________________________________
> Dev mailing list
> Dev at openlayers.org
> http://openlayers.org/mailman/listinfo/dev
>



More information about the Dev mailing list