On 2023/08/29 10:42, Jeff Law wrote:
>
>
> On 8/14/23 00:09, Tsukasa OI wrote:
>> From: Tsukasa OI <research_tra...@irq.a4lg.com>
>>
>> After commit c283c4774d1c ("RISC-V: Throw compilation error for unknown
>> extensions") changed how do we handle unknown extensions, we have no
>> guarantee that we can share the same architectural string with Binutils
>> (specifically, the assembler).
>>
>> To avoid compilation errors on shared Assembler-C/C++ projects, GCC
>> should
>> support almost all extensions that Binutils support, even if the GCC does
>> not touch a thing.
>>
>> This commit adds stub supported standard unprivileged extensions to
>> riscv_ext_version_table and its implications to riscv_implied_info
>> (all information is copied from Binutils' bfd/elfxx-riscv.c except not
>> yet
>> merged 'Zce', 'Zcmp' and 'Zcmt' support).
>>
>> gcc/ChangeLog:
>>
>> * common/config/riscv/riscv-common.cc
>> (riscv_implied_info): Add implications from unprivileged extensions.
>> (riscv_ext_version_table): Add stub support for all unprivileged
>> extensions supported by Binutils as well as 'Zce', 'Zcmp', 'Zcmt'.
>>
>> gcc/testsuite/ChangeLog:
>>
>> * gcc.target/riscv/predef-31.c: New test for a stub unprivileged
>> extension 'Zcb' with some implications.
> This series (most likely patch 3/3) seems to break arch-24.c and arch-25.c.
>
> Please fix and post a V3.
>
> Jeff
>
I think it was a hidden merge failure with partial Zc* extensions
support by Jiawei (and I already fixed it in the internal version).
I'll re-review it and submit as v3 if it's okay.
I don't recall exact test cases that failed (when I tested) but looking
at arch-24.c and arch-25.c you pointed out, they have a minor issue
(independent with this patch set). I'll submit a minor fix for those
files later.
Thanks,
Tsukasa
