[11u] Backport 8229352: Use of an uninitialized register in 32-bit ARM template interpreter

[Boris Ulasevich]

Hi Christoph,

I do not reproduce the CDS issue with ARM hard float. I think what you 
see is some kind of user error or a different problem.

Since tier1 is Ok we can ask jdk11u-fix-request for 8229352. I propose 
this message:

 > Fix Request
 > Patch fixes uninitialized register value error which makes 
interpreter to go though wrong path.
 > The fix is critical for performance and stability reasons.
 > Patch applies cleanly to 11u and passes tier1 on ARM sflt/hflt.

regards,
Boris

On 21.08.2019 14:26, christoph.goettschkes at microdoc.com wrote:
> Hi Boris,
> 
> I have the following issue:
> 
> The hotspot tier1 JTreg tests do not show any regresseion (as with the
> jdk/jdk repository), but the CDS archive tests are failing (with and
> without the patch applied). I guess this would block the backporting.
> Should I already send you the "Fix Request" comment and file a new bug
> report for that?
> 
> I did not see this issue when I tested the jdk/jdk repository (not sure if
> the issue is there too). I guess this is due to the fact that I changed
> the test setup (I have to use a different test device right now).
> 
> Not even the following test works:
> 
> $ ./jdk/bin/java -cp $(pwd) -XX:+UnlockDiagnosticVMOptions
> -XX:SharedArchiveFile=$(pwd)/cds.jsa -Xshare:dump -version
> 
> I am not able to reproduce this issue on amd64, but it is present on
> 32-bit ARM hard and soft float.
> 
> Thanks, Christoph