This is the mail archive of the mailing list for the GCC 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: stabs and #define/#undef

> Date: Tue, 24 Oct 2006 11:30:00 -0400
> From: Daniel Jacobowitz <>
> On Tue, Oct 24, 2006 at 07:59:01AM -0700, Ian Lance Taylor wrote:
> > Please send a patch to the gdb folks for gdb/doc/stabs.texinfo.  In
> > fact, please do that before you check the change into gcc, in case
> > they have any recommend changes.
> None of the GDB developers seem to care about stabs any longer, to be
> honest; our focus is mostly on modern platforms using DWARF-2.

Agreed.  By contrast EMC's primary interest is embedded.  And I
suspect that there are other embedded users out there that would like
to have some of the functionality that has been available for dwarf-2
for quite a while.

A problem arose internally that would be easily solved by having the
functionality.  I knew that the functionality was available for
dwarf-2, so I looked into what it would take to add it to stabs.  And
the answer was -- it's easy, the hard part is already done.  So, I did
it.  About 1 1/2 years ago.  But, we just recently got the paperwork
signed.  All past and future changes.  Now I've got several years
worth of changes to untangle and submit one by one...

Regarding dwarf-2 vs stabs -- there are two things standing in the way
of us converting from stabs to dwarf-2:

. internal scripts that process stabs info.  This is becoming less and
  less of an issue.  If the other issue was to go away *AND* a real
  benefit was seen to conversion (read: better debug experience), the
  time would be found to convert the scripts.

. the size of the dwarf-2 debug info relative to stabs.  The last time
  that I checked, dwarf-2 debug info was an order of magnitude larger
  than stabs.  Resulting in executables that were 2-3 times larger
  (and larger *.o's, too).  With a build being 12-15 GB, it adds up

> I don't know what issues there are with grabbing arbitrary numbers for
> the new stabs.  Other than that, I don't see what could be
> objectionable.

I look upon the stabs.texinfo and stabs.def files as the unofficial
``clearinghouse'' (for want of a better word) for stabs stuff.

> -- 
> Daniel Jacobowitz
> CodeSourcery

Thanks for your comments.

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