JDK-8275509: (jlink) SystemModulesPlugin generates a jdk.internal.module.SystemModules$all.class which isn't reproducible

[Alan Bateman]

On 19/10/2021 13:31, Jaikiran Pai wrote:
> This relates to the intermittent failures in 
> tools/jlink/JLinkReproducibleTest.java test case which has been 
> ProblemListed for a while now. The root cause is 
> https://bugs.openjdk.java.net/browse/JDK-8275509. I couldn't find any 
> specific mailing lists for jlink tool and I remember seeing 
> jlink/jpackage related discussions on this mailing list previously, so 
> creating this discussion here.

Reproducible builds have been a game of whack-a-mole. Many issues have 
been fixed in recent releases to the JDK is a lot better than it used to 
be. As it happens, someone else interested in reproducible builds 
brought up the issue of the hashCode on jigsaw-dev a few weeks ago [1].

The system modules plugin pre-dates the archiving of the module graph in 
the CDS archive so the performance profile is probably very different 
from when the plugin was created. I'll defer to Claes on what benchmarks 
to run.

As regards changing the plugin then if we aren't creating the MD with a 
hashCode then I'd prefer to drop the hashCode parameter from the Builder 
and the shared secret. In other words, do some cleanup rather than leave 
it orfaned.

-Alan

[1] 
https://mail.openjdk.java.net/pipermail/jigsaw-dev/2021-September/014708.html