Re: [OpenJDK 2D-Dev] RFR [11] JDK-8202580: Dashed BasicStroke randomly painted incorrectly, may freeze application

[Laurent Bourgès]

Thanks Phil & Sergey for your reviews.

I will push tomorrow morning (FR)...

Thanks Kevin for submitting the jfx11 bug.
FYI I am finalizing marlinFX 0.9.1 patch too (ClipShapeTest adapted)

Laurent

Le 7 mai 2018 9:20 PM, "Sergey Bylokhov"  a
écrit :

Looks fine.


On 07/05/2018 09:39, Philip Race wrote:
>> Sergey, what do you think ?

We can request a backport after the push to jdk11, if no new issues will
be found.



-- 
Best regards, Sergey.

Re: [OpenJDK 2D-Dev] RFR [11] JDK-8202580: Dashed BasicStroke randomly painted incorrectly, may freeze application

[Kevin Rushforth]

I filed https://bugs.openjdk.java.net/browse/JDK-8202743 to track 
getting this fix into JavaFX.


-- Kevin


On 5/4/2018 5:30 AM, Laurent Bourgès wrote:

Sergey,

Please review this simple fix to the Dasher problem:
JBS: https://bugs.openjdk.java.net/browse/JDK-8202580
webrev: http://cr.openjdk.java.net/~lbourges/marlin/marlin-8202580.0/ 



Changes:
- (D)Dasher.init: the given dash array is dirty as 
MarlinRenderingEngine got it from XxxArrayCache 
(Unsafe.allocateUninitializedArray) when calling copyDashArray(). The 
fix consists in using the correct part [0; dashLen[ as it was the case 
in jdk9
- added test that detects wrong dashed rectangle: jtreg fails on 
jdk10/11 without patch


Notes:
- the thread may hang in init() if the dirty part contains negative 
values (normalization loop never exits), but it is very difficult to 
reproduce and test.
- My Apologies: I introduced the bug in dec 2016 when I merged 
MarlinFX / Marlin2D.



Finally, the proposed fix should be backported to JDK10 and also to 
OpenJFX 10/11. I can propose a patch to OpenJFX11 soon.


Could somebody else manage the backport process (JDK-update 10, 
OpenJFX10) ?


Regards,
Laurent

Re: [OpenJDK 2D-Dev] RFR [11] JDK-8202580: Dashed BasicStroke randomly painted incorrectly, may freeze application

[Sergey Bylokhov]

Looks fine.

On 07/05/2018 09:39, Philip Race wrote:

Sergey, what do you think ?


We can request a backport after the push to jdk11, if no new issues will 
be found.



--
Best regards, Sergey.

Re: [OpenJDK 2D-Dev] RFR [11] JDK-8202580: Dashed BasicStroke randomly painted incorrectly, may freeze application

[Philip Race]

On 5/7/18, 9:27 AM, Laurent Bourgès wrote:

Phil,

Le dim. 6 mai 2018 à 01:11, Philip Race > a écrit :


Looks fine to me. Did you run the rendering related regression tests ?


I ran all tests in sun/java2d/marlin. What other tests should I run ?
This fix is very low risk as it fixes only the dash array usage.


That's OK. We'll run more tests in our nightly once it is pushed.



I am not sure if a backport to 10 will be approved under the new
guidelines :

http://mail.openjdk.java.net/pipermail/jdk-updates-dev/2018-April/000102.html


Agreed, it is not P1 but the reporter mentioned his swing app was 
hanging in some case at startup so this bug causes some stability 
problems.


It depends on your opinions if it is worth to be fixed asap or not.


Step 1 is to get it pushed to 11. It needs to bake before we can ask if 
we decide we should.


-phil.

Sergey, what do you think ?

Laurent

Re: [OpenJDK 2D-Dev] RFR [11] JDK-8202580: Dashed BasicStroke randomly painted incorrectly, may freeze application

[Laurent Bourgès]

Phil,

Le dim. 6 mai 2018 à 01:11, Philip Race  a écrit :

> Looks fine to me. Did you run the rendering related regression tests ?
>

I ran all tests in sun/java2d/marlin. What other tests should I run ?
This fix is very low risk as it fixes only the dash array usage.

>
> I am not sure if a backport to 10 will be approved under the new
> guidelines :
>
> http://mail.openjdk.java.net/pipermail/jdk-updates-dev/2018-April/000102.html


Agreed, it is not P1 but the reporter mentioned his swing app was hanging
in some case at startup so this bug causes some stability problems.

It depends on your opinions if it is worth to be fixed asap or not.
Sergey, what do you think ?

Laurent

Re: [OpenJDK 2D-Dev] RFR [11] JDK-8202580: Dashed BasicStroke randomly painted incorrectly, may freeze application

[Philip Race]

Looks fine to me. Did you run the rendering related regression tests ?

I am not sure if a backport to 10 will be approved under the new 
guidelines :

http://mail.openjdk.java.net/pipermail/jdk-updates-dev/2018-April/000102.html

-phil.

On 5/4/18, 5:30 AM, Laurent Bourgès wrote:

Sergey,

Please review this simple fix to the Dasher problem:
JBS: https://bugs.openjdk.java.net/browse/JDK-8202580
webrev: http://cr.openjdk.java.net/~lbourges/marlin/marlin-8202580.0/ 



Changes:
- (D)Dasher.init: the given dash array is dirty as 
MarlinRenderingEngine got it from XxxArrayCache 
(Unsafe.allocateUninitializedArray) when calling copyDashArray(). The 
fix consists in using the correct part [0; dashLen[ as it was the case 
in jdk9
- added test that detects wrong dashed rectangle: jtreg fails on 
jdk10/11 without patch


Notes:
- the thread may hang in init() if the dirty part contains negative 
values (normalization loop never exits), but it is very difficult to 
reproduce and test.
- My Apologies: I introduced the bug in dec 2016 when I merged 
MarlinFX / Marlin2D.



Finally, the proposed fix should be backported to JDK10 and also to 
OpenJFX 10/11. I can propose a patch to OpenJFX11 soon.


Could somebody else manage the backport process (JDK-update 10, 
OpenJFX10) ?


Regards,
Laurent