aarch64: Register clash in debug build

[Doerr, Martin]

Hi Andrew,

I've opened a bug for it. Feel free to take it (or somebody else):
https://bugs.openjdk.java.net/browse/JDK-8218670

Best regards,
Martin


-----Original Message-----
From: Doerr, Martin 
Sent: Montag, 4. Februar 2019 19:19
To: 'Andrew Haley' <aph at redhat.com>; hotspot-runtime-dev at openjdk.java.net; shade at redhat.com
Subject: RE: aarch64: Register clash in debug build

Hi Andrew,

we were running some jck tests with -Xcomp -XX:-TieredCompilation:
api/java_lang/constant/DynamicConstantDesc
api/java_lang/invoke/ConstantBootstraps
api/java_lang/invoke/MethodHandle
api/java_lang/reflect/Constructor/typeanno/getannotatedparametertypes

We got 5 hs_err files.

Best regards,
Martin


-----Original Message-----
From: Andrew Haley <aph at redhat.com> 
Sent: Montag, 4. Februar 2019 19:07
To: Doerr, Martin <martin.doerr at sap.com>; hotspot-runtime-dev at openjdk.java.net; shade at redhat.com
Subject: Re: aarch64: Register clash in debug build

On 2/4/19 2:47 PM, Doerr, Martin wrote:
> we had run -Xcomp tests on aarch64 and got SIGSEGV in StubRoutines::generic_arraycopy (debug build).
> 
> generate_generic_copy:
> rscratch2_dst_klass = rscratch2 = r9
> 
> It uses
> MacroAssembler::check_klass_subtype_slow_path
> which overwrites rscratch2 in debug build.
> 
> We don't support aarch64 at the moment, but I guess somebody is interesting in fixing it.

Thank you. Can you tell us how to reproduce this? Any particular test, or
will "Hello, World!" be enough?

-- 
Andrew Haley
Java Platform Lead Engineer
Red Hat UK Ltd. <https://www.redhat.com>
EAC8 43EB D3EF DB98 CC77 2FAD A5CD 6035 332F A671