[commit/AIX] Fix error happening while loading a core file

[commit/AIX] Fix error happening while loading a core file

[Joel Brobecker]

[-- Attachment #1: Type: text/plain, Size: 677 bytes --]

This is another regression that I noticed on AIX.

The problem is that we had a check in xcoff_relocate_symtab that
was supposed to detect the case where we're debugging a core file.
The check was based on the inferior ptid being null. I replaced it
with a check against target_has_execution, which seemed more
appropriate.

Is there a routine that performs this check?

2009-03-15  Joel Brobecker  <brobecker@adacore.com>

        Fix an error happening while loading symbols from a core file.

        * rs6000-nat.c (xcoff_relocate_symtab): Use target_has_execution
        to detect whether we're debugging a core file or not.

Tested on powerpc-aix. Checked in.

-- 
Joel

[-- Attachment #2: rs6000-nat.diff --]
[-- Type: text/x-diff, Size: 452 bytes --]

Index: rs6000-nat.c
===================================================================
--- rs6000-nat.c	(revision 146288)
+++ rs6000-nat.c	(working copy)
@@ -1036,7 +1036,8 @@ xcoff_relocate_symtab (unsigned int pid)
   int ldisize = arch64 ? sizeof (ldi->l64) : sizeof (ldi->l32);
   int size;
 
-  if (ptid_equal (inferior_ptid, null_ptid))
+  /* Nothing to do if we are debugging a core file.  */
+  if (!target_has_execution)
     return;
 
   do

Re: [commit/AIX] Fix error happening while loading a core file

[Pedro Alves]

On Sunday 15 March 2009 20:33:20, Joel Brobecker wrote:
> This is another regression that I noticed on AIX.
> 
> The problem is that we had a check in xcoff_relocate_symtab that
> was supposed to detect the case where we're debugging a core file.
> The check was based on the inferior ptid being null. 

I see, no multi-threaded/lwps core files on aix then, as
otherwise this would have already been noticed before.

> I replaced it 
> with a check against target_has_execution, which seemed more
> appropriate.
> 
> Is there a routine that performs this check?

Nope, that's what we've got.  We'll have to do something about these
target_has_* for multi-process, but meanwhile, that's indeed the
correct test.

-- 
Pedro Alves