[cairo] Endless loop in _cairo_stroker_line_to_dashed()
chris at chris-wilson.co.uk
Sun Jun 28 14:07:40 PDT 2009
On Sun, 2009-06-28 at 19:01 +0200, Hans Breuer wrote:
> At 28.06.2009 13:32, Hans Breuer wrote:
> > At 28.06.2009 12:24, Chris Wilson wrote:
> >> It would be good to have a test case for this bug.
> > Yes, definitely.
> >> (Or you can provide instructions on how I can
> >> reproduce this bug myself using dia, and I'll create a test case.)
> Dia 0.96.1 (only PDF), 0.97 or current master should all work to reproduce.
> > Unfortunately I was not yet able to create a further reduced version of the
> > proprietary file which still enters the endless loop. But maybe the
> > above is enough?
> Now I have a dia file with only two boxes and a line which makes cairo go
> into that endless loop. I'm attaching the cairo-script file created with
> Dia's new cairo-script export facility ;)
Eeek! A real user of the CairoScript backend, I'd better make sure it
still works. ;-)
> But I wonder if the precision given by that file is enough to reproduce the
> bug. And I didn't find a way to check myself. csi-replay.c looked
> promising, but I can't find a binary created from it.
Possibly not, Joonas was complaining about the limited precision earlier
and I promised him I would implement doubles as part of the binary
csi-replay should be sufficient to execute those scripts, but is not
built by default. So you need to "make -C util/cairo-script csi-replay".
Though in this case I think csi-exec would be more useful as that would
generate a png directly. However, the reason why they are no longer
built is that they have been superseded by test/cairo-test-trace and
perf/cairo-perf-trace. The intention is to install these alongside cairo
once the utilities have matured sufficiently to be broadly useful.
Thank you for the script - at first glance it doesn't appear to do
anything different than Joonas' test case, but it is worth a closer
More information about the cairo