Re: [PATCH] elf.h: Drop duplicate DT_RISCV_VARIANT_CC

[Fangrui Song <maskray@google.com>]

On 2023-04-28, Palmer Dabbelt wrote:
>I missed the update from last year that included these already, having
>duplicates breaks the build due to redefinition errors like
>
>    elf/elf.h:4013: error: macro STO_RISCV_VARIANT_CC redefined
>    elf.h:3941: note: location of previous definition
>
>Reported-by: Joseph S. Myers <joseph@codesourcery.com>
>Fixes: 117e8b341c ("riscv: Resolve symbols directly for symbols with STO_RISCV_VARIANT_CC.")
>Link: https://inbox.sourceware.org/libc-alpha/mhng-0d9fb5a0-63fa-4b02-8029-7c20232f39ee@palmer-ri-x1c9/T/#t
>Signed-off-by: Palmer Dabbelt <palmer@rivosinc.com>
>---
>bulid-many-glibcs is still churning and might take a bit, but looks like
>this was just a silly mistake on my end.  I'm pretty confident this
>fixes the issue, but happy to wait for the tests to finish given that I
>just screwed this up.  That was before coffee, though, so maybe I can
>avoid some of the blame... ;)
>---
> elf/elf.h | 7 -------
> 1 file changed, 7 deletions(-)
>
>diff --git a/elf/elf.h b/elf/elf.h
>index 4f65b5a32d..94ca23c1bb 100644
>--- a/elf/elf.h
>+++ b/elf/elf.h
>@@ -3933,13 +3933,6 @@ enum
>
> #define R_TILEGX_NUM		130
>
>-/* RISC-V specific values for the Dyn d_tag field.  */
>-#define DT_RISCV_VARIANT_CC	(DT_LOPROC + 1)
>-#define DT_RISCV_NUM		2

We probably should keep DT_RISCV_NUM. There is just one definition.

>-/* RISC-V specific values for the st_other field.  */
>-#define STO_RISCV_VARIANT_CC 0x80

The convention appears to place DT_ STO_ definitions after R_
definitions for a particular processor.
I'd suggest that we move the definitions below R_ while keeping just one
copy.

> /* RISC-V ELF Flags */
> #define EF_RISCV_RVC 			0x0001
> #define EF_RISCV_FLOAT_ABI 		0x0006
>-- 
>2.40.0
>