[PATCH 3/3] RISC-V: Changed confused conflict message of zcmp

[Nelson Chu]

Oops, thanks for reminding me of this history.  It's just that
-march=rv64id_zcmp doesn't cause the conflict, so I'm a bit confused...  I
think I will change the message to,

% riscv64-unknown-elf-as -march=rv64i_zcd_zcmp empty.s -o empty.o
Assembler messages:
Error: zcmp' is incompatible with `d' and `c', or `zcd' extension

riscv64-unknown-elf-as -march=rv64idc_zcmp empty.s -o empty.o
Assembler messages:
Error: zcmp' is incompatible with `d' and `c', or `zcd' extension

So that the error message should be the same as
riscv_multi_subset_supports_ext.  Committed in the previous patch.

Thanks
Nelson

On Wed, Jul 9, 2025 at 5:04 PM Jiawei <jiawei@iscas.ac.cn> wrote:

>
> 在 2025/7/9 13:25, Jan Beulich 写道:
> > On 09.07.2025 06:53, Nelson Chu wrote:
> >> Not sure why the message said "zcmp' is incompatible with `d/zcd'
> extension",
> >> but we only check if zcd exsit or not.
> > I expect the message was that way as a consequence of the C+D effects
> > that were there even before Zcd was invented. As Zcd has been around
> > for quite some time, modernizing the message is likely fine.
> >
> > Jan
>
> Yes, the message wants to means Zcmp is both conflict with Zcd and D
> extension.
>
> Since it use some encodings from c.fsdsp and it depands on Zca extension.
>
> Once you have D, it will turn Zcd on, so marked d/zcd here.
>
>
> It's okay to change this if you feel confused about this.
>
>
> BR,
>
> Jiawei
>
> >> --- a/bfd/elfxx-riscv.c
> >> +++ b/bfd/elfxx-riscv.c
> >> @@ -2223,7 +2223,7 @@ riscv_parse_check_conflicts (riscv_parse_subset_t
> *rps)
> >>         && riscv_subset_supports (rps, "zcd"))
> >>       {
> >>         rps->error_handler
> >> -    (_("zcmp' is incompatible with `d/zcd' extension"));
> >> +    (_("zcmp' is incompatible with `zcd' extension"));
> >>         no_conflict = false;
> >>       }
> >>     if (riscv_lookup_subset (rps->subset_list, "zcf", &subset)
>
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <https://sourceware.org/pipermail/binutils/attachments/20250710/1b4e3b1c/attachment.htm>