Re: [RFA] [patch] 'info symbol' to print more info

[Paul Pluzhnikov <ppluzhnikov@google.com>]

On Fri, Nov 14, 2008 at 1:44 PM, Michael Snyder <msnyder@vmware.com> wrote:

> Just one suggestion (and suggestion only) -- since the exec_file
> is sort of the default/common case, do you think it would be a
> good idea to check if it's the exec_file, and omit the objfile
> information if so?

I think there are two large classes of users:
- embedded people with no shared libraries and a single (statically
  linked) executable, and
- everybody else.

I've never been in the first camp. From the second camp, if the
symbol is in the main executable, I want to know that too.

E.g. if I do 'info symbol &malloc', and it tells me .text in a.out,
that may well be a very important clue (that something unusual is
happening).

I do agree that for the fully-static executable case the extra
verbiage may be somewhat annoying. Looking for a way to tell if
that is the case ... Would the test below be sufficient?

  if (ojbect_files->next)
   /* there is only one object, so don't print its name ... */

I also just noticed inconsistency between 'info symbol' and
'maintenance translate-address' wording:

  exit+0 section .text in /usr/lib64/libc.so.6
  exit in section .text of /usr/lib64/libc.so.6

I'll fix that in the final patch so they are the same.

Thanks,
-- 
Paul Pluzhnikov