Possible stabs bugs

Possible stabs bugs

[Al Stevens]

I'm writing a GUI debugger for the egcs 1.1 port of Mingw32. Mumit suggested
I submit bugs here. Is there a news group that I should be monitoring?

1. cc1plus.exe emits a strange stab for this code:

struct { int x, y; } foo;

If you name the struct as shown next, the stab looks right.

struct bar { int x, y; } foo;


2. The linker is not resolving the addresses of global variables (N_GSYM) in
the stab as previous versions of gnu-win32 do. The egcs 1.1 linker leaves
the stab's value field zero and places no other symbolic infomation in the
exe file that (as far as I can tell) provides the variable's memory address.

3. The stabs format has changed considerably since the 1993 FSF spec by
Menapace, Kingdon, and MacKenzie. Can you direct me to a more recent
specification document?

Many thanks.

Al Stevens
Dr. Dobb's Journal

Re: Possible stabs bugs

[Mumit Khan]

On 11 Sep 1998, Stan Shebs wrote:

> alstevens@midifitz.com (Al Stevens) writes:
> 
> > 2. The linker is not resolving the addresses of global variables (N_GSYM) in
> > the stab as previous versions of gnu-win32 do. The egcs 1.1 linker leaves
> > the stab's value field zero and places no other symbolic infomation in the
> > exe file that (as far as I can tell) provides the variable's memory address.
> 
> Interesting, but what is an "egcs 1.1 linker"?

I believe Al is talking about the linker in my egcs-1.1 binary 
distribution for cygwin32/mingw32. Until recently, the packaged
binutils was v2.9.1 (plus quite a few changes), and for 1.1, I
switched over to 980830 dev snapshot which fixed lots of annoying
PE bugs. I've added some local modifications, but those only affect 
windres, dlltool, etc, and not bfd nor the linker.

Regards,
Mumit

Re: Possible stabs bugs

[Jeffrey A Law]

  In message < xdk93aft7j.fsf@andros.cygnus.com >you write:
  > > 3. The stabs format has changed considerably since the 1993 FSF spec by
  > > Menapace, Kingdon, and MacKenzie. Can you direct me to a more recent
  > > specification document?
  > 
  > Changed?  In what way?  We update the stabs document as changes occur,
  > but they've been very infrequent.  I'd certainly like to hear about
  > any undocumented changes.
This could be the changes to support BINCL which are documented in
the current stabs docs, but I do not think they were in the circa
1993 stabs docs.

jeff