RFR: Loom: Add missing stubs to fix build failures on PPC64, S390, ARM32 and Zero [v3]

[Alan Bateman]

On Fri, 28 Jan 2022 06:58:13 GMT, Hao Sun <haosun at openjdk.org> wrote:

>> PPC64, S390, ARM32, Zero stubs were initially introduced to Loom in
>> commit 3a89cd5 [1]. However, necessary updates were missing afterward,
>> leading to compilation failures.
>> 
>> 1. frame_helpers.inline.hpp
>> 
>> In commit 854a81a [2], platform specific implementation was extracted
>> from continuation.cpp to frame_helpers.inline.hpp, and the corresponding
>> headers for only x86 and aarch64 were created then.
>> 
>> This same updates should be done to PPC64/S390/ARM32/Zero as well.
>> Otherwise, the following compilation error would be emitted.
>> 
>> 
>> ~/jdk_src/src/hotspot/share/runtime/frame_helpers.inline.hpp:326:1:
>> fatal error: frame_helpers_ppc.inline.hpp: No such file or directory
>> 
>>   326 | #include CPU_HEADER_INLINE(frame_helpers)
>> 
>>         | ^~~~~~~~~~~~~~~~~
>> 
>>         compilation terminated.
>> 
>> 
>> Fix: see the updates in continuation_XX.inline.hpp and
>> frame_helpers_XX.inline.hpp.
>> 
>> 2. ZGC support for PPC64
>> 
>> ZGC support for Loom was implemented two years ago in commit 05b1b53
>> [3]. Recently ZGC for PPC64 was supported in the mainline. See commit
>> 337b73a [4].
>> 
>> However, stub `BarrierSetNMethod::arm()` is missing for PPC64 platform.
>> 
>> Here is the snippet of error log when cross-compiling on PPC64 in my
>> local test environment.
>> 
>> 
>> * For target
>> * hotspot_variant-server_libjvm_gtest_objs_BUILD_GTEST_LIBJVM_link:
>
> Hao Sun has updated the pull request incrementally with one additional commit since the last revision:
> 
>   Fix compilation warning in ARM32 build
>   
>   The following error message would be emitted for ARM32 build.
>   
>   ```
>   /home/runner/work/loom/loom/jdk/src/hotspot/share/oops/instanceStackChunkKlass.cpp:283:28: error: format '%ld' expects argument of type 'long int', but argument 5 has type 'intptr_t' {aka 'int'} [-Werror=format=]
>     283 |       assert (offset >= 0, "Derived pointer offset is %ld", offset);
>         |                            ^~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~  ~~~~~~
>         |                                                             |
>      ... (rest of output omitted)
>   ```
>   
>   The fix is easy, i.e. using INTPTR_FORMAT.

Marked as reviewed by alanb (Committer).

-------------

PR: https://git.openjdk.java.net/loom/pull/87