[PATCH, MIPS] Better diagnostic on incorrect size for .reginfo

[Maciej W. Rozycki]

On Wed, 15 Aug 2007, Adam Nemet wrote:

> For $newabi targets I am forcing n32 now.  This seems to be in sync
> with current practice (e.g. reloc-1-n32).  Maciej, Is this close
> enough to what you had in mind?

 Yes, but please see the notes below.

> --- ld-mips-elf/mips-elf.exp	13 Aug 2007 21:16:39 -0000	1.47
> +++ ld-mips-elf/mips-elf.exp	15 Aug 2007 19:02:33 -0000
> @@ -297,3 +297,10 @@ run_dump_test "attr-gnu-4-31"
>  run_dump_test "attr-gnu-4-32"
>  run_dump_test "attr-gnu-4-33"
>  run_dump_test "attr-gnu-4-41"
> +
> +# N64 doesn't have .reginfo, force n32.
> +if $has_newabi {
> +    run_dump_test "reginfo-size-n32"
> +} else {
> +    run_dump_test "reginfo-size"
> +}

 I think you should drop the "else" clause and run the "reginfo-size" test 
unconditionally.

> --- /dev/null	1 Jan 1970 00:00:00 -0000
> +++ ld-mips-elf/reginfo-size.d	15 Aug 2007 19:02:33 -0000
> @@ -0,0 +1,3 @@
> +#source: reginfo-size.s
> +#ld: -Treginfo-size.ld
> +#error: .reginfo is .* bytes instead of .* bytes

 And add "-mabi=32" here.

 I think the mips64*-*-* target should test as much of the o32 ABI 
functionality as possible so that subtle bugs based on configuration are 
not accidentally introduced (well, they will regardless, but the denser 
the testing sieve, the fewer will get through).

  Maciej