Why not a label for .debug_frame?

Why not a label for .debug_frame?

[Stan Shebs]

I've been looking into adopting Dwarf 2 for Darwin / Mac OS X, and
as part of that, tried enabling -gdwarf-2 in GCC's Darwin port.

However, I have a problem where the name of the section .debug_frame
is used as a symbol (a "section symbol"), which is an ELFism not
supported by the Mach-O assembler.  I'm a little loath to change
the assembler to create section symbols, and have been looking at
how the .debug_frame symbol is used.  In the process I noticed that
other sections such as .debug_info and .debug_abbrev have labels
generated for them, and subsequent code refers to those labels
rather than the sections.  (To help jog memories, the patch was
from Robert Lipe for SVR5, and may be seen at
http://gcc.gnu.org/ml/gcc-patches/1999-02n/msg00802.html .)

So my question is, why isn't .debug_frame handled this way too?
Inertia perhaps?

Stan

Re: Why not a label for .debug_frame?

[Richard Henderson]

On Mon, Sep 24, 2001 at 05:35:38PM -0700, Stan Shebs wrote:
> So my question is, why isn't .debug_frame handled this way too?

Oversight?


r~

Re: Why not a label for .debug_frame?

[Robert Lipe]

Stan Shebs wrote:
> I've been looking into adopting Dwarf 2 for Darwin / Mac OS X, and
> as part of that, tried enabling -gdwarf-2 in GCC's Darwin port.

As a maintainer, you're unlikely to regret it. 

> However, I have a problem where the name of the section .debug_frame
> is used as a symbol (a "section symbol"), which is an ELFism not
> supported by the Mach-O assembler.  I'm a little loath to change

I'm not convinced it's an ELF-ism at all.  I think it's something that
just happens to be supported by GNU as.

> the assembler to create section symbols, and have been looking at
> how the .debug_frame symbol is used.  In the process I noticed that
> other sections such as .debug_info and .debug_abbrev have labels
> generated for them, and subsequent code refers to those labels
> rather than the sections.  (To help jog memories, the patch was
> from Robert Lipe for SVR5, and may be seen at
> http://gcc.gnu.org/ml/gcc-patches/1999-02n/msg00802.html .)
> 
> So my question is, why isn't .debug_frame handled this way too?
> Inertia perhaps?

I selfishly scratched only the rash that was affecting me.  I was in
code that was well above my head.  I beat it until submission so that my
target bootstrapped and the debugger worked at least trivially.  So I
made no promises that I took the moral high roads on that.   

RJL

Re: Why not a label for .debug_frame?

[Stan Shebs]

Robert Lipe wrote:
> 
> > However, I have a problem where the name of the section .debug_frame
> > is used as a symbol (a "section symbol"), which is an ELFism not
> > supported by the Mach-O assembler.  I'm a little loath to change
> 
> I'm not convinced it's an ELF-ism at all.  I think it's something that
> just happens to be supported by GNU as.

Yeah, somehow I always "just knew" this, but I couldn't find any
documented reference to the behavior, and finally had to verify by
looking at gas sources...

> > So my question is, why isn't .debug_frame handled this way too?
> > Inertia perhaps?
> 
> I selfishly scratched only the rash that was affecting me.  I was in
> code that was well above my head.  I beat it until submission so that my
> target bootstrapped and the debugger worked at least trivially.  So I
> made no promises that I took the moral high roads on that.

No prob, now that I have a similar itch to scratch (image of
hundreds of engineers all furiously scratching themselves, bleah),
I'll work up the patch.

Stan