On Sat, 14 Jan 2023 07:23:18 GMT, Sergey Bylokhov <[email protected]> wrote:

> Currently, we always validate image layouts passed from java to native, 
> except for some types of rasters. One of the reasons why the rasters are 
> skipped is that they are not public and have their own validation during 
> construction. But it is unclear how it works in the multithreaded case. All 
> these objects are not thread-safe - could it be possible that we can get the 
> raster that is not completely constructed/validated?
> 
> FYI I have started to look into this method because when I did an accidental 
> typo in pixel format in this method I got the same symptoms/crash as 
> described [here](https://bugs.openjdk.org/browse/JDK-8278101). It might be 
> unrelated but it is safer to fix.

This pull request has now been integrated.

Changeset: 7348b9ec
Author:    Sergey Bylokhov <[email protected]>
URL:       
https://git.openjdk.org/jdk/commit/7348b9ec9373746bb76bc9fa9556f1811bd9e475
Stats:     9 lines in 1 file changed: 5 ins; 1 del; 3 mod

8300167: Add validation of the raster's layout before using in native

Reviewed-by: prr

-------------

PR: https://git.openjdk.org/jdk/pull/12004

Reply via email to