Re: [patch] Replace reread_symbols by load+free calls

[Pedro Alves <pedro at codesourcery dot com>]

On Thursday 25 June 2009 20:21:20, Jan Kratochvil wrote:
> + ? ? ?for (so = master_so_list (); so != NULL; so = so->next)
> +???????if (so->objfile == objfile)
> +??????? ?break;
> + ? ? ?if (so != NULL)
> +???????continue;
> ?

Would this work instead?

if ((objfile->flags & (OBJF_SHARED | OBJF_USERLOADED) == OBJF_SHARED))
  continue;

The OBJF_USERLOADED check is to take care of the windows-nat.c
"dll-symbols" command...

Referencing to master_so_list outside of the solib files doesn't sound
right.


On Thursday 25 June 2009 20:21:20, Jan Kratochvil wrote:
> +      /* FIXME: EXEC_BFD should be always referenced only through
> +        SYMFILE_OBJFILE.  Currently exec_bfd exists on its own and it does not
> +        have to be listed among OBJECT_FILES.  exec_bfd_mtime should either be
> +        used or rather dropped.  */

Can you elaborate on this FIXME note?  exec_bfd may or not have symbols in it.
It may or not be the same bfd as symfile_objfile.  Note that there are
"exec-file" and "symbol-file" separate commands.  In fact, I believe,
historicaly, the "file" command that does both things at once was only
added later.  Do you plan on somehow eliminating exec_bfd?

>    if (reread_one)
>      {
> +      /* breakpoint_re_set gets called to comply with SYMFILE_DEFER_BP_RESET
> +        above.  */
>        clear_symtab_users ();

I can't quite parse what this comment is really saying.  Who calls
breakpoint_re_set?  Who is complying?

-- 
Pedro Alves