Re: Need Help in using GDB

[Bob Proulx]

Chiranjaya Tantrige wrote:
> We are having some issues in using gdb. When we use gdb to read a core file
> it gives below highlighted error. Please advice on how to resolve it.

What highlighted error?  How did you highlight it?

Also you should know that mailing lists messages are always going to
be plain text only.  If your message was mime multipart alternative
with a text plain part then the text plain part will be used by the
mailing list.  Please only send plain text email.

These messages caught my eye:

> warning: Can't read pathname for load map: I/O error.
> ...
> warning: Can't read pathname for load map: I/O error.
> warning: Can't read pathname for load map: I/O error.

Because the reported error is an I/O error I suspect this isn't a
software problem but is instead a hardware problem.  An I/O error is
being reported to gdb and gdb is reporting it up the chain.  Please
check the system log (/var/log/syslog? probably named different on
your sparc system) and look for any system reported I/O errors.

If you find something suspicious then I would try cat'ing the file to
/dev/null.  ('cat foo > /dev/null' as a test.) If you can't read the
file then gdb can't read the file either.

However!  I also see these messages when I search the archives.

  [patch] Fix GNU/Linux core open: Can't read pathname for load map: 
Input/output error.
  http://sourceware.org/ml/gdb-patches/2009-10/msg00142.html
  http://sourceware.org/ml/gdb-patches/2011-05/msg00372.html

That sounds exactly like the problem you are seeing.  I know nothing
more about it.

> Reading symbols from
> /u01/xxx/corefiles/TestCorefiles_xxx/Corefiles_100.11/AuthServer...done

I imagine this is an NFS mounted directory.  If you can't find
anything that appears to be a hardware I/O error from the system, and
if it doesn't appear to be the other problem either, then I would
secondly suspect some interaction problem with the networked file
system.

I would try to make a local copy of all of the needed files.  If you
can make everything available in a local disk location, say
/tmp/mydebug or some such, then it would eliminate that interaction
from the debug session.  It might work then.  Don't know.  Just
guessing.  But have seen problems similar to that fail over NFS and
work on the local disk before.  Maybe it would for you too.

Unfortunately from me that is about as much help as I know to give on
this topic.

Bob