This is the mail archive of the gdb-patches@sourceware.org mailing list for the GDB project.


Index Nav: [Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav: [Date Prev] [Date Next] [Thread Prev] [Thread Next]
Other format: [Raw text]

Re: [rfc] Remove current_gdbarch register handling from symbol readers


Eli Zaretskii wrote:
> > Date: Thu, 21 May 2009 02:52:22 +0200 (CEST)
> > From: "Ulrich Weigand" <uweigand@de.ibm.com>
> > Cc: gdb-patches@sourceware.org
> > 
> > Only the Windows/DOS related targets remain.  Of those, all the
> > -pe / -mingw / -cygwin / -interix targets actually use the PE
> > file format, not "real" COFF, and GCC will default to generate
> > stabs or DWARF-2 for those.  It seems possible to use -gcoff to
> > force GCC to emit COFF/SDB debug info; I'm not sure if this will
> > actually work with PE files ...
> 
> Well, I already tried that, and at least with MinGW GCC 3.4.2 that I
> have installed, it fails: the binary produced with -gcoff is not
> recognized by MinGW GDB 6.8 (it says "not in executable format").
> Maybe it's a bug somewhere.
> 
> > The only target for which GCC by default generates COFF/SDB debug
> > info does indeed appear to be -msdosdjgpp ...
> 
> DJGPP GCC uses DWARF-2 by default also, but it fully supports -gcoff.

Ah, OK, I have misread the GCC config files ...

So overall it seems that *no* target uses COFF/SDB debug info as its
default format any more; those that still support it at all also use
a more powerful format as default.  Under those circumstances, is there
any benefit to continuing to support that format in GDB, given that it
becomes more and more diffcult to test that this support actually
still works?
 
> The problem is that DJGPP does not support expect, so the the test
> suite cannot be run.  I can run some tests manually, so if you or
> someone else could tell which test may be affected by this change, I
> can try running them.

Pretty much any access to an in-register local variable or parameter
should break if the change is wrong.  Thanks for your help in testing!

Bye,
Ulrich

-- 
  Dr. Ulrich Weigand
  GNU Toolchain for Linux on System z and Cell BE
  Ulrich.Weigand@de.ibm.com


Index Nav: [Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav: [Date Prev] [Date Next] [Thread Prev] [Thread Next]