Re: recent MIPS mdebug-related changes cause gcc testsuitefailures.

[cgd at broadcom dot com]

At Fri, 21 Jun 2002 14:54:19 -0400, Daniel Jacobowitz wrote:
> That's what I meant (well, should have meant at least) by the second
> half.  Telling the assembler -g is wrong too, for non-assembly source.

But... how do you distinguish asm vs. non-asm source?

i mean, if you say:

mips-elf-gcc -gcoff foo.S

i believe the intended effect is, after preprocessing, to run the
assembler with flags that include "-g" and now "-mdebug", right?

in this case, you _do_ want assembler debugging info, right?

I.e., I don't think it's Right to get rid of most of the
SUBTARGET_ASM_DEBUGGING_SPEC contents...


> Fixing this is a little hard, because it requires actually checking for
> the (recent, GAS only) -mdebug option.  You should probably add a
> configure check for it rather than trying to key it off MASK_GAS, since
> mips-elf doesn't even default to GAS.  Just copy the check for
> --gstabs in configure.in and pick an appropriate version number (maybe
> --2.13).
> 
> Then you do:
> 
> #if HAVE_AS_MDEBUG_DEBUG_FLAG
> #define SUBTARGET_ASM_DEBUGGING_SPEC "%{gcoff* -mdebug}"
> #else
> ...
> #endif
> 
> ASM_DEBUG_SPEC needs to be adjusted also, probably, to pass -mdebug;
> but that's not important.

Uh... I don't see an ASM_DEBUG_SPEC...

Personally, I'd think that handling -mdebug (${mdebug} would belong in
SUBTARGET_ASM_DEBUGGING_SPEC along with the rest, since it _is_ an
assembler debugging flag...


chris