handle halfway case in scan converter

handle halfway case in scan converter

Scan edges that start at exactly -0.5 aren't trimmed by
clipping or by rounding, triggering a debug assert.

One way to fix this is to round the top and left
down instead of up.

Also, move the path initialization of gm/composeshader.cpp
to make debugging other path problems easier.

R=reed@google.com
BUG=skia:2715
GOLD_TRYBOT_URL= https://gold.skia.org/search2?unt=true&query=source_type%3Dgm&master=false&issue=1544873002

Committed: https://skia.googlesource.com/skia/+/6df611574a3cf8abf2617af0d03a5553bb17360d

[caryclark, 2015-12-22 19:24:59]

Description was changed from

==========
handle halfway case in scan converter

Scan edges that start at exactly -0.5 aren't trimmed by
clipping or by rounding, triggering a debug assert.

One way to fix this is to round the top and left
down instead of up.

Also, move the path initialization of gm/composeshader.cpp
to make debugging other path problems easier.

R=reed@google.com
BUG=skia:2715
==========

to

==========
handle halfway case in scan converter

Scan edges that start at exactly -0.5 aren't trimmed by
clipping or by rounding, triggering a debug assert.

One way to fix this is to round the top and left
down instead of up.

Also, move the path initialization of gm/composeshader.cpp
to make debugging other path problems easier.

R=reed@google.com
BUG=skia:2715
GOLD_TRYBOT_URL=
https://gold.skia.org/search2?unt=true&query=source_type%3Dgm&master=false&is...
==========

[caryclark, 2015-12-22 19:40:11]

My fix may be the wrong approach -- couldn't figure out how to change the scan
converter itself. Anyways, there's a nice test case here that reproduces the
bug.

[reed1, 2016-01-04 17:49:38]

Lets try to make dround() itself just a local helper function at the call-site
(only seems to be one in SkScan_Path.cpp:585). Then we don't have to modify
either public header.

[caryclark, 2016-01-04 19:24:10]

On 2016/01/04 17:49:38, reed1 wrote:
> Lets try to make dround() itself just a local helper function at the call-site
> (only seems to be one in SkScan_Path.cpp:585). Then we don't have to modify
> either public header.

Done.

[caryclark, 2016-01-04 19:24:19]

The CQ bit was checked by caryclark@google.com to run a CQ dry run

[commit-bot: I haz the power, 2016-01-04 19:24:34]

Dry run: CQ is trying da patch. Follow status at
 https://chromium-cq-status.appspot.com/patch-status/1544873002/60001
View timeline at
 https://chromium-cq-status.appspot.com/patch-timeline/1544873002/60001

[commit-bot: I haz the power, 2016-01-04 19:42:48]

The CQ bit was unchecked by commit-bot@chromium.org

[commit-bot: I haz the power, 2016-01-04 19:42:49]

Dry run: This issue passed the CQ dry run.

[reed1, 2016-01-04 19:43:39]

https://codereview.chromium.org/1544873002/diff/60001/src/core/SkScan_Path.cpp
File src/core/SkScan_Path.cpp (right):

https://codereview.chromium.org/1544873002/diff/60001/src/core/SkScan_Path.cp...
src/core/SkScan_Path.cpp:581: *      SkScalar x = 0.49999997f;
I think the comment block needs to be expanded, now that we're treating top/left
differently from bottom/right

https://codereview.chromium.org/1544873002/diff/60001/src/core/SkScan_Path.cp...
src/core/SkScan_Path.cpp:587: static void round_out(const SkRect& src, SkIRect*
dst) {
SkRect uses "out" to mean we floor/ceiling the values. Perhaps we need a name
more specific... round_asymmetric_to_int?

[caryclark, 2016-01-04 20:58:42]

https://codereview.chromium.org/1544873002/diff/60001/src/core/SkScan_Path.cpp
File src/core/SkScan_Path.cpp (right):

https://codereview.chromium.org/1544873002/diff/60001/src/core/SkScan_Path.cp...
src/core/SkScan_Path.cpp:581: *      SkScalar x = 0.49999997f;
On 2016/01/04 19:43:39, reed1 wrote:
> I think the comment block needs to be expanded, now that we're treating
top/left
> differently from bottom/right

Done.

https://codereview.chromium.org/1544873002/diff/60001/src/core/SkScan_Path.cp...
src/core/SkScan_Path.cpp:587: static void round_out(const SkRect& src, SkIRect*
dst) {
On 2016/01/04 19:43:39, reed1 wrote:
> SkRect uses "out" to mean we floor/ceiling the values. Perhaps we need a name
> more specific... round_asymmetric_to_int?

Done.

[reed1, 2016-01-04 22:00:09]

lgtm

[caryclark, 2016-01-04 22:02:53]

The CQ bit was checked by caryclark@google.com

[commit-bot: I haz the power, 2016-01-04 22:03:00]

CQ is trying da patch. Follow status at
 https://chromium-cq-status.appspot.com/patch-status/1544873002/80001
View timeline at
 https://chromium-cq-status.appspot.com/patch-timeline/1544873002/80001

[commit-bot: I haz the power, 2016-01-04 22:17:48]

Description was changed from

==========
handle halfway case in scan converter

Scan edges that start at exactly -0.5 aren't trimmed by
clipping or by rounding, triggering a debug assert.

One way to fix this is to round the top and left
down instead of up.

Also, move the path initialization of gm/composeshader.cpp
to make debugging other path problems easier.

R=reed@google.com
BUG=skia:2715
GOLD_TRYBOT_URL=
https://gold.skia.org/search2?unt=true&query=source_type%3Dgm&master=false&is...
==========

to

==========
handle halfway case in scan converter

Scan edges that start at exactly -0.5 aren't trimmed by
clipping or by rounding, triggering a debug assert.

One way to fix this is to round the top and left
down instead of up.

Also, move the path initialization of gm/composeshader.cpp
to make debugging other path problems easier.

R=reed@google.com
BUG=skia:2715
GOLD_TRYBOT_URL=
https://gold.skia.org/search2?unt=true&query=source_type%3Dgm&master=false&is...

Committed:
https://skia.googlesource.com/skia/+/6df611574a3cf8abf2617af0d03a5553bb17360d
==========

[commit-bot: I haz the power, 2016-01-04 22:17:53]

Committed patchset #5 (id:80001) as
https://skia.googlesource.com/skia/+/6df611574a3cf8abf2617af0d03a5553bb17360d