[patch] python: load *-gdb.py for shlibs during attach

[patch] python: load *-gdb.py for shlibs during attach

[Jan Kratochvil]

Hi,

currently if you attach to or load a core file for executable, its -gdb.py
script is loaded.  But none -gdb.py for the shared libraries.

Spawned or with the fix (libstdc++-gdb.py):
#1  0x00000000004007b5 in f (s="a") at 1.C:4
                             ^ is std::string

Attached/core without the fix:
#1  0x00000000004007b5 in f (s=...) at 1.C:4

No regressions on {x86_64,x86_64-m32,i686}-fedora13-linux-gnu.

Mostly obvious, I will check it in later.


Thanks,
Jan


gdb/
2010-09-22  Jan Kratochvil  <jan.kratochvil@redhat.com>

	* main.c: Include objfiles.h.
	(captured_main): New variable objfile.  Call
	load_auto_scripts_for_objfile for ALL_OBJFILES.

--- ./gdb/main.c	2010-09-22 10:51:32.000000000 +0200
+++ ./gdb/main.c	2010-09-22 10:50:44.000000000 +0200
@@ -43,6 +43,7 @@
 #include "source.h"
 #include "cli/cli-cmds.h"
 #include "python/python.h"
+#include "objfiles.h"
 
 /* The selected interpreter.  This will be used as a set command
    variable, so it should always be malloc'ed - since
@@ -323,6 +324,7 @@ captured_main (void *data)
 
   int i;
   int save_auto_load;
+  struct objfile *objfile;
 
   struct cleanup *pre_stat_chain = make_command_stats_cleanup (0);
 
@@ -957,8 +959,8 @@ Can't attach to process and specify a co
      We wait until now because it is common to add to the source search
      path in local_gdbinit.  */
   gdbpy_global_auto_load = save_auto_load;
-  if (symfile_objfile != NULL)
-    load_auto_scripts_for_objfile (symfile_objfile);
+  ALL_OBJFILES (objfile)
+    load_auto_scripts_for_objfile (objfile);
 
   for (i = 0; i < ncmd; i++)
     {

Re: [patch] python: load *-gdb.py for shlibs during attach

[Tom Tromey]

>>>>> "Jan" == Jan Kratochvil <jan.kratochvil@redhat.com> writes:

Jan> currently if you attach to or load a core file for executable, its
Jan> -gdb.py script is loaded.  But none -gdb.py for the shared
Jan> libraries.

Jan> 2010-09-22  Jan Kratochvil  <jan.kratochvil@redhat.com>
Jan> 	* main.c: Include objfiles.h.
Jan> 	(captured_main): New variable objfile.  Call
Jan> 	load_auto_scripts_for_objfile for ALL_OBJFILES.

FWIW, I agree this is the right thing to do.

Tom

Re: [patch] python: load *-gdb.py for shlibs during attach

[Jan Kratochvil]

On Wed, 22 Sep 2010 21:23:00 +0200, Tom Tromey wrote:
> Jan> 2010-09-22  Jan Kratochvil  <jan.kratochvil@redhat.com>
> Jan> 	* main.c: Include objfiles.h.
> Jan> 	(captured_main): New variable objfile.  Call
> Jan> 	load_auto_scripts_for_objfile for ALL_OBJFILES.
> 
> FWIW, I agree this is the right thing to do.

BTW what do you think about the case when the called python scripts would
manipulate the ALL_OBJFILES list?  But even ALL_OBJFILES_SAFE would not be
safe enough in such case.

Sure it could be protected some other way like to lookup again the saved
objfile name.  But it can also deadlock.  I guess there are more places like
that in GDB.


Regards,
Jan

Re: [patch] python: load *-gdb.py for shlibs during attach

[Tom Tromey]

>>>>> "Jan" == Jan Kratochvil <jan.kratochvil@redhat.com> writes:

Jan> BTW what do you think about the case when the called python scripts would
Jan> manipulate the ALL_OBJFILES list?  But even ALL_OBJFILES_SAFE would not be
Jan> safe enough in such case.

I'm inclined not to worry about it.

I guess we could come up with a scheme to make it work.  That seems like
a lot of effort for something that, at present, is just a theoretical
problem.

Tom

Re: [patch] python: load *-gdb.py for shlibs during attach

[Jan Kratochvil]

On Wed, 22 Sep 2010 21:41:19 +0200, Tom Tromey wrote:
> I'm inclined not to worry about it.

Checked-in:
	http://sourceware.org/ml/gdb-cvs/2010-09/msg00136.html


Thanks,
Jan