[private] Re: [External] : Re: [Bug] javac gives non-deterministic output (for sealed interface with records)

[Vicente Romero]

Hi Przemek,

The fix to the issue you reported [1] has been integrated in our master 
repo. Thanks a lot for all the help tracking this very interesting issue 
down,

Thanks,
Vicente

[1] https://bugs.openjdk.org/browse/JDK-8322477
[2] https://github.com/openjdk/jdk

On 1/9/24 11:04, Vicente Romero wrote:
> Hi Przemek,
>
> this is pretty good, thanks for the reproducer. I will update the bug 
> entry,
>
> Thanks,
> Vicente
>
> On 1/9/24 03:48, Przemek Bielicki wrote:
>> Hey Vicente,
>>
>> My fellow Gradle JVM Platform engineer (Octavia Togami) created a 
>> reproducer: 
>> https://github.com/octylFractal/javac-permitted-ordering-bug 
>> <https://urldefense.com/v3/__https://github.com/octylFractal/javac-permitted-ordering-bug__;!!ACWV5N9M2RV99hQ!OuIjNEls9z95A5Eq458IZEjWn1nlxh3v2_wq3jYrPj61R62YNl6fUCJ_-Docq9rmUcpTVMWgk3iQLhvjhTxyZA$>
>> Her first analysis was: "Looking at the stacktraces, it appears that 
>> when it's ordered properly, the subclasses of the sealed type are 
>> visited as part of their parent; but when it's ordered incorrectly, 
>> they're visited as part of resolving the type of a variable somewhere 
>> else."
>> Eventually she says: "This is not in any way a Gradle bug. It has to 
>> do with ordering of other files that use the sealed type and its 
>> subclasses."
>>
>> Can you please verify this on your end and update the JIRA ticket 
>> with the latest findings?
>>
>> Thanks,
>> Przemek
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <https://mail.openjdk.org/pipermail/compiler-dev/attachments/20240110/81334edf/attachment-0001.htm>