RFR: 8275509: ModuleDescriptor.hashCode isn't reproducible across builds [v4]

Jaikiran Pai jai.forums2013 at gmail.com
Mon Nov 1 02:14:55 UTC 2021


On 01/11/21 7:26 am, Jaikiran Pai wrote:
> Hello Alan,
>
> On 01/11/21 1:03 am, Alan Bateman wrote:
>> On Fri, 29 Oct 2021 04:19:21 GMT, Jaikiran Pai <jpai at openjdk.org> wrote:
>>
>>> Do you mean the jtreg `driver` style this test is using? I used it 
>>> because it was necessary to compare output (the hashCode value) 
>>> between multiple JVM runs. I use the `driver` along with the 
>>> `ProcessBuilder` to capture the output between multiple JVM runs and 
>>> compare those.
>> No, I mean the test needs cleanup so it can easily be maintained. For 
>> starters I think the main method can take a parameter to indicate its 
>> the child process. HashCodeChecker.main becomes a method in 
>> ModuleDescriptorHashCodeTest.  You'll see several examples of this in 
>> the test suite.
>
> I looked around some existing tests and I could find some which have a 
> "realMain" and a "main" methods. But in all such test cases, the 
> "realMain" is just called as a method within the same process. So I 
> guess that's not the one you meant. I'll look around some more to try 
> and find the expected style for such child processes.

I think I might have understood what you meant. I have now updated the 
PR to restructure the test case accordingly. Please let me know if this 
looks fine or if this isn't what you meant.

-Jaikiran



More information about the core-libs-dev mailing list