Re: [PATCH v2] RISC-V: hash with segment id and pcrel_hi address while recording pcrel_hi

[Nelson Chu <nelson@rivosinc.com>]

[-- Attachment #1: Type: text/plain, Size: 963 bytes --]

The const asection * sounds reasonable and good to me, thanks for the
suggestion :-)

Nelson

On Thu, Jul 4, 2024 at 2:46 PM Jan Beulich <jbeulich@suse.com> wrote:

> On 04.07.2024 03:56, lifang_xia@linux.alibaba.com wrote:
> > --- a/gas/config/tc-riscv.c
> > +++ b/gas/config/tc-riscv.c
> > @@ -1784,6 +1784,7 @@ init_opcode_hash (const struct riscv_opcode
> *opcodes,
> >     help us resolve the corresponding low-part relocation later.  */
> >  typedef struct
> >  {
> > +  segT seg;
>
> Just one minor comment here: segT really being asection *, I think it would
> be desirable to use const asection * here instead, to clarify that there's
> no intent of changing the referenced section. Not using the pre-cooked segT
> may not be very nice, but (a) there's precedent (obj-elf.c:match_section())
> and (b) I think const-correctness is more important overall.
>
> In the end it'll be Nelson anyway to approve the change.
>
> Jan
>