[RFA/commit] fix problem when debugging a core file on Tru64

[Joel Brobecker]

Hello,

Core file debugging on Tru64 is slightly broken:

    (gdb) core core
    Reading symbols from /sbin/loader...done.
    Loaded symbols for /sbin/loader
    Reading symbols from /usr/shlib/libc.so...done.
    Loaded symbols for /usr/shlib/libc.so
    You can't do that without a process to debug.

The problem comes again from the solib create_inferior_hook which tries
to resume the execution of the inferior.  In the case of debugging
a core file, this cannot (and doesn't need to) be done.

I fixed it by checking whether the target_can_run or not. It seemed
logical to check that target feature before trying to resume the inferior.
However, it introduces another use of the global current_target. I can't
remember if we are trying to eliminate this global or not...

2008-06-02  Joel Brobecker  <brobecker@adacore.com>

        * solib-osf.c (osf_solib_create_inferior_hook): Do nothing if
        the target cannot run.

Tested on alpha-tru64. Will commit in a week if no comment.

-- 
Joel
-------------- next part --------------
Modified: solib-osf.c
===================================================================
--- solib-osf.c	2008-06-02 16:51:00 UTC (rev 132042)
+++ solib-osf.c	2008-06-02 17:00:22 UTC (rev 132043)
@@ -321,7 +321,14 @@
   /* Now run the target.  It will eventually get a SIGTRAP, at
      which point all of the libraries will have been mapped in and we
      can go groveling around in the rld structures to find
-     out what we need to know about them. */
+     out what we need to know about them.
+     
+     If debugging from a core file, we cannot resume the execution
+     of the inferior.  But this is actually not an issue, because
+     shared libraries have already been mapped anyways, which means
+     we have nothing more to do.  */
+  if (!target_can_run (&current_target))
+    return;
 
   clear_proceed_status ();
   stop_soon = STOP_QUIETLY;