Sunday, June 29, 2025

Re: [Question] databases/db/v4,-java on aarch64

On Jun 28, 2025, at 3:36 PM, Stuart Henderson <stu@spacehopper.org> wrote:
>
> On 2025/06/28 21:46, Slava Voronzoff wrote:
>> Hello, I tried today to launch libreoffice on my aarch64 machine and got
>>
>> Warning: failed to launch javaldx - java may not function correctly
>> Unspecified Application Error
>>
>> after trying to install libreoffice-java I noticed that db-java not existed.
>>
>> Little investigation and I found this
>>
>> ONLY_FOR_ARCHS-java = amd64 i386
>> # XXX aarch64 disabled until --execute-only situation can be resolved in jdk
>> #ONLY_FOR_ARCHS-java = aarch64 amd64 i386
>>
>> I enabled aarch64, built it and succesfully start lowriter and more or less use it (editing/saving).
>>
>> So my question is:
>>
>> what was this --execute-only situation and isnt it resolved yet?
>
> cc'ing kurt - it may have been resolved in this commit
>
> ---------------------
> PatchSet 526
> Date: 2023/02/03 20:32:57
> Author: kurt
> Branch: HEAD
> Tag: (none)
> Log:
> * Remove errant aarch64 hotspot assembly instruction that was reading from
> libjvm.so .text segment. This is clearly not was intended here. Found by
> exec only.
> * Remove USE_NOEXECONLY flags.
> * Disable -Werror for 11 & 17 to so clang 15 can build them.
>
> Members:
> 1.8/Makefile:1.77->1.78
> 1.8/patches/patch-hotspot_src_cpu_aarch64_vm_templateInterpreter_aarch64_cpp:INITIAL->1.1
> 11/Makefile:1.40->1.41
> 11/patches/patch-make_common_NativeCompilation_gmk:1.5->1.6
> 11/patches/patch-src_hotspot_cpu_aarch64_templateInterpreterGenerator_aarch64_cpp:1.2->1.3
> 17/Makefile:1.14->1.15
> 17/patches/patch-make_common_NativeCompilation_gmk:1.2->1.3
> 17/patches/patch-src_hotspot_cpu_aarch64_templateInterpreterGenerator_aarch64_cpp:INITIAL->1.1
> ---------------------
>
>> Maybe it is ok to enable aarch64 back on db4-java ?
>
> looks like this might be ok indeed.

Yes. This was fixed with that commit. Since then I have run the java test
kit on aarch64 which does thousands of tests on aarch64. As far as I know
it is EXECONLY clean unless a new case has crept in since my testing.

Best,
-Kurt

No comments:

Post a Comment