mis-elf-as : local/global symbol order in symbol table

[Pierre Habraken]

Ian Lance Taylor wrote:
> Pierre Habraken <Pierre.Habraken@imag.fr> writes:
> 
>> The ELF specification states that table all local symbols in a symbol
>> should precede all global symbols.
>> mips-elf-as (binutils 2.20) disregards this rule.
>> Can anyone explain to me the reason for this ?
> 
> It's a historical issue.  The SGI Irix 5 assembler did not follow this
> rule, and the Irix 5 linker got confused when gas did try to follow
> this rule.  As I recall, Irix 5 thought the rule applied to
> STT_SECTION symbols rather than to STB_LOCAL symbols.  To work around
> this we (or maybe just me) introduced all the elf_bad_symtab code.
> Later this was extended to the IRIX_COMPAT and SGI_COMPAT code.  We
> can probably straighten out mips-elf-as if we try.

Hello Ian,

Thanks for your reply.
I do not know what are "elf_bad_symtab", "IRIX_COMPAT" and "SGI_COMPAT". 
I assume that they are code chunks in binutils, are they ?

I have to precise that I am not a mips programmer. I just supervise a 
student project in which students have to program an assembler or a 
simulator. The mips architecture was choosen for educational reasons 
(not for technical reasons). Several students ask us why the "official" 
assembler (mips-elf-as !) does not conform to the ELF specification 
(this is rather confusing for them). Hence my post.

Pierre
-- 
Pierre HABRAKEN - mailto:Pierre.Habraken@imag.fr
Tél: 04 76 63 59 94 - Fax: 04 76 51 47 93