Bug 62375

Created attachment 76567 [details] [review]
Simplify get-path-extents.c so that only this specific bug is tested (might make it easier to figure out what is going wrong)

Created attachment 76662 [details] [review]
Another patch which makes the test case succeed and I still don't know what I am doing

Also please notice that there are still differences between the old code and the current one, they generate different values for "stop".

This bug appears to be the only thing keeping me from setting up automated regular runs of make test.  Which I would really like to be doing.

diff --git a/src/cairo-gstate.c b/src/cairo-gstate.c
index 6319471..b8caa63 100644
--- a/src/cairo-gstate.c
+++ b/src/cairo-gstate.c
@@ -1462,19 +1462,19 @@ _cairo_gstate_stroke_extents (cairo_gstate_t	 *gstate,
     }
 
     if (status == CAIRO_INT_STATUS_UNSUPPORTED) {
-	cairo_traps_t traps;
-
-	_cairo_traps_init (&traps);
-	status = _cairo_path_fixed_stroke_polygon_to_traps (path,
-							    &gstate->stroke_style,
-							    &gstate->ctm,
-							    &gstate->ctm_inverse,
-							    gstate->tolerance,
-							    &traps);
-	empty = traps.num_traps == 0;
+	cairo_polygon_t polygon;
+
+	_cairo_polygon_init (&polygon, NULL, 0);
+	status = _cairo_path_fixed_stroke_to_polygon (path,
+						      &gstate->stroke_style,
+						      &gstate->ctm,
+						      &gstate->ctm_inverse,
+						      gstate->tolerance,
+						      &polygon);
+	empty = polygon.num_edges == 0;
 	if (! empty)
-	    _cairo_traps_extents (&traps, &extents);
-	_cairo_traps_fini (&traps);
+	    extents = polygon.extents;
+	_cairo_polygon_fini (&polygon);
     }
     if (! empty) {
 	_cairo_gstate_extents_to_user_rectangle (gstate, &extents,

That last patch does cause get-path-extents tests to pass for me.  Can you commit it so I can start running this from cron?

490 Passed, 2 Failed [0 crashed, 2 expected], 32 Skipped
PASS: cairo-test-suite

With --enable-gl --enable-xcb --disable-ps --disable-script, CAIRO_REF_DIR, and using xvfb.

We (packagers) would love to run tests for cairo too, should we backport that patch (for an undefined amount of time) or will someone push it “soon”?

commit 9ea5993b036f5930179263baaf3162eeebb7c153
Author: Chris Wilson <chris@chris-wilson.co.uk>
Date:   Thu Jun 20 14:11:38 2013 +0100

    test/get-path-extents: Check exact matches within tolerance
    
    When we refine geometry, we do so to a tolerance as specified by the
    user. This means that we can not expect tessellated results to have
    exact results, but always they should match within the specified
    tolerance.
    
    Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>

commit b7331f0c52cc64f2c224eac502afa6c50a1a8d8b
Author: Chris Wilson <chris@chris-wilson.co.uk>
Date:   Thu Jun 20 14:23:15 2013 +0100

    gstate: Speed up stroked path extents
    
    We can skip the intermediate evaluation of the trapezoids for
    determining the extents of a stroked path by using the relatively new
    functions for computing the contours of the stroke. Then we can simply
    use the bbox of the points within the contours to retrieve the path
    extents - which is already provided by the polygon holding the contours
    of the stroke. This provides a faster result with less numerical
    inaccuracy due to fewer stages required in the computation
    
    References: https://bugs.freedesktop.org/show_bug.cgi?id=62375
    Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>

Verified, thanks:  http://lists.cairographics.org/archives/cairo/2013-June/024411.html