[cairo] [PATCH] Fix intersect_with_boxes() to produce tight clip extents

[Andrea Canciani]

On Fri, Jun 24, 2011 at 5:39 PM, Uli Schlachter <psychon at znc.in> wrote:
> On 21.06.2011 10:19, Taekyun Kim wrote:
>> From: Taekyun Kim <tkq.kim at samsung.com>
>>
>> Previous code was intersecting extents with infinitely large rectangle.
>> ---
>>  src/cairo-clip.c |    4 ++--
>>  1 files changed, 2 insertions(+), 2 deletions(-)
>>
>> diff --git a/src/cairo-clip.c b/src/cairo-clip.c
>> index a695cd7..e272d54 100644
>> --- a/src/cairo-clip.c
>> +++ b/src/cairo-clip.c
>> @@ -1399,8 +1399,8 @@ intersect_with_boxes (cairo_composite_rectangles_t *extents,
>>      cairo_rectangle_int_t rect;
>>      cairo_box_t box;
>>
>> -    box.p1.x = box.p1.y = INT_MIN;
>> -    box.p2.x = box.p2.y = INT_MAX;
>> +    box.p1.x = box.p1.y = INT_MAX;
>> +    box.p2.x = box.p2.y = INT_MIN;
>>      while (num_boxes--) {
>>       if (boxes->p1.x < box.p1.x)
>>           box.p1.x = boxes->p1.x;
>
> Hi,
>
> This patch broke the xcb backend. It looks like it is clipping away too much,
> because some of the resulting graphics are missing. Also, most of the tests that
> broke are about clipping. I bisected with clip-disjoint clip-fill-eo-unbounded
> trap-clip, see below.

I believe that the issue you are seeing might be related with:
https://bugs.freedesktop.org/show_bug.cgi?id=38641

Andrea

>
> clip-disjoint only looks broken on xcb-render-0.0. The other two tests broke on
> all backends that I have enabled, except ps.
>
> I don't have much time to investigate right now, but since even image is
> affected, perhaps someone else can figure this out.
>
> Cheers,
> Uli
>
>
> $ git bisect log
> git bisect start
> # bad: [77b6563c218dcff641820d2555ac93ad10a095e8] xcb: Use defines instead of
> magic numbers
> git bisect bad 77b6563c218dcff641820d2555ac93ad10a095e8
> # good: [f409f74decf63b3dbf68b713317a1eb17cd71a0f] xlib: Fix build of xlib-xcb
> git bisect good f409f74decf63b3dbf68b713317a1eb17cd71a0f
> # good: [b6d593dc71e276576ce2af9fa2267eebdddd578a] build: Clean up environment
> variables and flags for the win32 build
> git bisect good b6d593dc71e276576ce2af9fa2267eebdddd578a
> # bad: [5291f7ccad54b587bd1bee3a89426cf1d158c5bb] script-interpreter: Build on win32
> git bisect bad 5291f7ccad54b587bd1bee3a89426cf1d158c5bb
> # bad: [441f9c5037dd32464bd87e21108b702c0a3c508a] Fix intersect_with_boxes() to
> produce tight clip extents
> git bisect bad 441f9c5037dd32464bd87e21108b702c0a3c508a
> # good: [8fb9a6727e2b769bfa388c6027ebad0a53ffa02b] boilerplate: Use correct flag
> type in DFBWindowDescription
> git bisect good 8fb9a6727e2b769bfa388c6027ebad0a53ffa02b
>
> $ git bisect good
> 441f9c5037dd32464bd87e21108b702c0a3c508a is the first bad commit
> commit 441f9c5037dd32464bd87e21108b702c0a3c508a
> Author: Taekyun Kim <tkq.kim at samsung.com>
> Date:   Tue Jun 21 17:19:58 2011 +0900
>
>    Fix intersect_with_boxes() to produce tight clip extents
>
>    Previous code was intersecting extents with infinitely large rectangle.
>
>    Signed-off-by: Chris Wilson <chris at chris-wilson.co.uk>
>
> :040000 040000 6c050d8078f84e94c5b14fbd8538e628dc7a394f
> 501f64c104551fd540d5487372e46d817a12c1a4 M      src
>
> --
> - Buck, when, exactly, did you lose your mind?
> - Three months ago. I woke up one morning married to a pineapple.
>  An ugly pineapple... But I loved her.
> --
> cairo mailing list
> cairo at cairographics.org
> http://lists.cairographics.org/mailman/listinfo/cairo
>