Subject: Re: CVS commit: src/gnu/dist/gcc/gcc/config/i386
To: Krister Walfridsson <>
From: Matthias Drochner <>
List: tech-toolchain
Date: 07/19/2004 21:56:03 said:
> GDB have problems with stabs; I don't think I have managed to debug
> any non-trivial program on NetBSD 2.0 without GDB crashing or
> misbehaving in other ways.

Just for curiosity - does gdb6 work better, or don't they care
about stabs either?

> And you really want dwarf when debugging C++ too.
> [...]

Actually, whenever I tried to debug something in-depth I came
to a point where gdb was useless. Didn't cope with threads, didn't
show useful stackframes, couldn't singlestep or whatever. I always
ended up with printf()s or other code instrumentation...
(That's not to argue about debug symbol formats, just a side note.
It might explain however why almost noone cares about debugging

As I said in another mail, when I changed gcc-3.3's default in -current,
I primaliry wanted to establish something which can be pulled up to
the 2.0 branch or not, depending on further development. At that point
I was still optimistic that the dwarf2 output got fixed in the gcc-3.3
Now Christos had this change pulled up to the branch. For some
reason supposedly...

> I think that gcc must be changed back to use dwarf.

Agreed, it should be changed back. This is where gcc develops,
stabs is more or less historical.
I'll do so if noone beats me in this. (and noone will notice:-(
For the 2.0 branch, I'd like to hear Christos' opinion, and
other toolchain gurus.

In the medium term, we should face a switch to gcc-3.4 which is
much better in this respect. This needs to be done platform-by-
platform probably, so we need some infrastructure in gnu/dist
and *.mk...

best regards