openjdk team mailing list archive

[tony mancill <tmancill@xxxxxxxxxx>]

On Fri, Sep 25, 2020 at 03:51:38PM +0100, Julian Gilbey wrote:
> On Thu, Sep 24, 2020 at 05:21:46PM -0700, tony mancill wrote:
> > On Thu, Sep 24, 2020 at 10:52:50AM +0100, Julian Gilbey wrote:
> > > Great, and there's a bug report #944738
> > > (https://bugs.debian.org/cgi-bin/bugreport.cgi?bug=944738) for
> > > openjdk-11; I haven't checked if there are bugs reported against the
> > > other two versions.
> > 
> > I'm not able to reproduce the problem with openjdk-15.  Perhaps the
> > non-determinism has already been addressed upstream?
> 
> Oh, interesting.  So the non-determinism problem still exists, and
> java.base.jmod contains non-deterministic SHA-256 hashes.  But jlink
> seems to no longer check whether the hashes match, so the jlink tests
> now work.  This does seem to be to be a poor resolution to the
> (Debian) bug!  No idea why they've done this, though.

Thank you for looking into this.  That's obviously a short-coming of the
test I have been using.

> So I suggest that we still use the patch with openjdk-15 to avoid
> introducing non-determinism into the build, but this is clearly less
> important than on 11 and 14.

Makes sense.  This will prevent breakage later on if the jlink behavior
reverts to checking the hash.

I had already prepared the NMU for openjdk-15 - building first without
the patch and then with it so I could test with both - so I will go
ahead and upload that package today.

> > openjdk-13 does have the problem, so we could do another upload, but
> > given that it's not an LTS release and 14 is already available and
> > patched, do we need to?
> 
> Good question.  It's easy to do, and it does fix a bug, so I don't see
> a good reason not to do so.

Agreed.  (I could make a quip about my laptop SSD endurance, which is
visibly decreasing, but only about 1% per month.)

Seriously, I want to mention that I am willing to help get the NMU
changes committed to the OpenJDK team repo if the team is open to that.

> > Neither 13 nor 15 have open bugs filed against them for the jlink hash
> > issue.
> 
> Indeed, but we can still fix it (and in the changelog say something
> like "fixes bug #... for this version of openjdk").

Will do.  I was just noting it since you noted it earlier in the bug
thread.

Thanks,
tony

Attachment: signature.asc
Description: PGP signature