Re: MIPS_STABS_ELF is now broken

[Daniel Jacobowitz <drow at mvista dot com>]

On Thu, Jun 13, 2002 at 09:04:02AM -0700, H . J . Lu wrote:
> On Thu, Jun 13, 2002 at 11:49:11AM -0400, Daniel Jacobowitz wrote:
> > > 
> > > I am having the second thought. The gas change is too big for me. May
> > > I propose to do these:
> > > 
> > > 1. Modify the mips gas to take either
> > > 
> > > 	.file "foo.c"
> > > or
> > > 
> > > 	.file 1 "foo.c"
> > > 
> > > as the first line for stabs, dwarf and mdebug.
> > > 
> > > 2. Modify gcc 3 to emit
> > > 
> > > 	.file "foo.c"
> > > 
> > > as the first line for stabs, dwarf and mdebug.
> > > 
> > > In this way, the new gas will still work with the existing compiler
> > > and the dwarf/stabs output will be very similar to x86. That means you
> > > have to use the new gas for gcc 3. I will post my gas/gcc changes
> > > shortly.
> > 
> > What advantage do you claim this has over the patches I posted (which
> 
> My gas change will be very small and the dwarf/stabs asm output will be
> very similar to x86, which is well tested.

That really doesn't have any relevance.  MIPS and x86 obviously do not
expect the same sort of debug information, either at assembly or binary
level, or we wouldn't be here.  We are much closer to Alpha, the other
.mdebug-supporting target; we are also now behaving like Alpha, after
my patches.

> > have already gone in on the GCC side, and which do not make GCC3
> > require a new binutils version for stabs output)?
> 
> What is wrong to require a new binutils for a new gcc? Also if you do
> 
> # .../configure
> # make bootstrap
> 
> you will need a new binutils for mips anyway.

Didn't we just have this conversation in the opposite direction?

> > Also, your patch will be incompatible with traditional MIPS assemblers
> > and their use of .file.  All the world is not GNU as!
> > 
> 
> You haven't seen my gcc patch. How do you know it isn't compatible
> with non gas?

Sure, you can conditionalize things specially for GAS in this regard; I
did the same thing in the patch that went in yesterday.  But there's no
reason to support a syntax incompatible with all existing MIPS
compilers and all existing MIPS assemblers.

-- 
Daniel Jacobowitz                           Carnegie Mellon University
MontaVista Software                         Debian GNU/Linux Developer