This is the mail archive of the
gdb@sources.redhat.com
mailing list for the GDB project.
Re: New object file parser?
- From: Jim Blandy <jimb at redhat dot com>
- To: Dan Towner <dant at picochip dot com>
- Cc: gdb at sources dot redhat dot com
- Date: 12 Jul 2002 11:50:40 -0500
- Subject: Re: New object file parser?
- References: <3D2EFDE0.159BCF83@picochip.com>
Dan Towner <dant@picochip.com> writes:
> I am porting gdb to a new processor architecture (embedded
> multi-processor DSP). Currently, although we use gcc as our compiler, we
> have a proprietary assembler and linker, which generate non-standard
> object and executable files. Now, faced with the task of getting the
> symbol table information from the object file through to gdb, is it
> easier to:
>
> 1) Write a new parser for BFD.
>
> or
>
> 2) Write a converter which generates a standard object file format from
> our proprietary format.
Well, you've made your bed... :)
> Whichever route I take, which object file would be the easier to deal
> with? From what I have seen, I think that a.out would be the best, but
> is there anything simpler?
Well, there's S-records. I don't know if they can carry symbol
information, though. I think not.
If you need a symbol table, I'd say a.out is probably the simplest one
there is. You can probably find documentation for a.out in the SunOS
4 man pages, or perhaps somewhere on the Web.
The next question is, what debugging format do you use? Converting
debugging info from the proprietary format to STABS or Dwarf is going
to be the hard part, I'd bet.