[Bug gold/21285] R_386_GOT32 for baseless case looks implemented incorrectly.

[hjl.tools at gmail dot com]

https://sourceware.org/bugzilla/show_bug.cgi?id=21285

--- Comment #3 from H.J. Lu <hjl.tools at gmail dot com> ---
(In reply to Cary Coutant from comment #2)
> OK, suppose we have the following sequence of instructions:
> 
>   cmp  $0x8b, %al
>   movl address@hidden, %eax
>   mov  address@hidden(%ecx), %esp
> 
> which assemble to:
> 
>    0: 3c 8b                   cmp    $0x8b,%al
>    2: a1 00 00 00 00          mov    0x0,%eax
>    7: 8b a1 00 00 00 00       mov    0x0(%ecx),%esp
> 
> How is the linker supposed to distinguish the two mov instructions? We
> cannot make this optimization based on looking at the opcodes -- we need
> separate relocations to tell us which is which, unless it's OK to invalidate
> 0xa1 as a valid modrm byte in this usage.
> 
> Doesn't ld have the same problem?

movl address@hidden, %eax

is only supported with the updated i386 psABI, which requires
"mov address@hidden, %eax must be encoded with opcode 0x8b, not 0xa0,
to allow linker optimization."

-- 
You are receiving this mail because:
You are on the CC list for the bug.