This is the mail archive of the gdb@sources.redhat.com 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: [5.3] Alpha and that mdebug problem?


On Tue, 19 Nov 2002 19:07:34 -0500, Andrew Cagney <ac131313@redhat.com> said:

> Has anyone tried using the 5.3 branch of GDB on an alpha?  Bug
> reports, such as:
> http://sources.redhat.com/cgi-bin/gnatsweb.pl?cmd=edit&database=gdb&pr=676
> keep comming back to haunt poor GDB.  You'll note that they
> reproduced the problem with a current GDB.

I really wish somebody would do something about mdebugread.c: it's the
only debug format that doesn't use buildsym.c, and it's the only place
where GDB uses sorted linear blocks to store symbols.  I've already
had one patch shelved because it would have slowed down GDB's
performance under that debug format; given the presence of PR gdb/676,
though, for all I know that debug format is already currently broken.
To be honest, I wasn't too mad that the patch in question was shelved
- it was only a cleanup patch, rather than one that actually improved
GDB's performance - but I will be annoyed if future, useful patches of
mine get rejected because they might conceivably hurt a debug format
that nobody is willing to support.

And no, I'm not willing to support it myself, either: if I had a
little more free time and access to a relevant machine, I might give
it a try, but neither of those is the case right now.  But, given the
presence of PR gdb/676, I would prefer that, for now, we should adopt
a policy towards mdebugread.c that changes that don't knowingly cause
GDB to completely stop working on those targets are acceptable.

Incidentally, the specific reason that I was given for not making
mdebugread.c worse was that:

> However backwards we may find them, people are still using mdebug,
> and making the test suite unusable would make it impossible for a
> company like Red Hat to give them a code drop when they need one.

If it really is true that Red Hat wishes to be able to support people
using that debug format, perhaps somebody at Red Hat should spend some
time buildsym-ifying it and fixing PR gdb/676.

David Carlton
carlton@math.stanford.edu


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