[patch] Error reporting during insert_breakpoints

[patch] Error reporting during insert_breakpoints

[Michael Snyder]

[-- Attachment #1: Type: text/plain, Size: 354 bytes --]

Two small changes

1) If insert_breakpoints fails during stepping, "keep_going()"
contains a TRY_CATCH block, but does not print the exception error.
If the same thing happens during continue, "catch_errors" reports it.

2) insert_bp_locations() generates error text that assumes
memory breakpoints.  Confusing if remote_insert_breakpoints
is using Z0.


[-- Attachment #2: breaks.txt --]
[-- Type: text/plain, Size: 1477 bytes --]

2009-06-03  Michael Snyder  <msnyder@vmware.com>

	* infrun.c (keep_going): If insert_breakpoints fails, 
	print the reason.
	* breakpoint.c (insert_bp_location): Don't assume memory
	breakpoints when printing error message.

Index: infrun.c
===================================================================
RCS file: /cvs/src/src/gdb/infrun.c,v
retrieving revision 1.384
diff -u -p -r1.384 infrun.c
--- infrun.c	3 Jun 2009 18:50:36 -0000	1.384
+++ infrun.c	3 Jun 2009 20:33:40 -0000
@@ -4247,6 +4247,7 @@ keep_going (struct execution_control_sta
 	    }
 	  if (e.reason < 0)
 	    {
+	      exception_print (gdb_stderr, e);
 	      stop_stepping (ecs);
 	      return;
 	    }
Index: breakpoint.c
===================================================================
RCS file: /cvs/src/src/gdb/breakpoint.c,v
retrieving revision 1.401
diff -u -p -r1.401 breakpoint.c
--- breakpoint.c	3 Jun 2009 18:16:43 -0000	1.401
+++ breakpoint.c	3 Jun 2009 20:33:40 -0000
@@ -1223,10 +1223,10 @@ Note: automatically using hardware break
 	      else
 		{
 		  fprintf_unfiltered (tmp_error_stream, 
-				      "Cannot insert breakpoint %d.\n", 
+				      "Cannot insert breakpoint %d", 
 				      bpt->owner->number);
 		  fprintf_filtered (tmp_error_stream, 
-				    "Error accessing memory address ");
+				    " at memory address ");
 		  fputs_filtered (paddress (bpt->address), tmp_error_stream);
 		  fprintf_filtered (tmp_error_stream, ": %s.\n",
 				    safe_strerror (val));

Re: [patch] Error reporting during insert_breakpoints

[Pedro Alves]

On Wednesday 03 June 2009 21:42:14, Michael Snyder wrote:
> Two small changes
> 
> 1) If insert_breakpoints fails during stepping, "keep_going()"
> contains a TRY_CATCH block, but does not print the exception error.
> If the same thing happens during continue, "catch_errors" reports it.

Yes, please.  This silent stop bites me often, when longjmp-resume
breakpoints fail to insert due to glibc pointer mangling...

> 2) insert_bp_locations() generates error text that assumes
> memory breakpoints.  Confusing if remote_insert_breakpoints
> is using Z0.

Strictly speaking, Z0 breakpoints *are* supposed to have
memory breakpoints semantics.  :-)

 @item z0,@var{addr},@var{length}
 @itemx Z0,@var{addr},@var{length}
 @cindex @samp{z0} packet
 @cindex @samp{Z0} packet
 Insert (@samp{Z0}) or remove (@samp{z0}) a memory breakpoint at address
 @var{addr} of size @var{length}.

> @@ -1223,10 +1223,10 @@ Note: automatically using hardware break
>               else
>                 {
>                   fprintf_unfiltered (tmp_error_stream, 
> -                                     "Cannot insert breakpoint %d.\n", 
> +                                     "Cannot insert breakpoint %d", 
>                                       bpt->owner->number);
>                   fprintf_filtered (tmp_error_stream, 
> -                                   "Error accessing memory address ");
> +                                   " at memory address ");
>                   fputs_filtered (paddress (bpt->address), tmp_error_stream);
>                   fprintf_filtered (tmp_error_stream, ": %s.\n",
>                                     safe_strerror (val));

I'm not objecting to the change, but, honestly, I don't see where
the assumption was, or where it is being removed?

-- 
Pedro Alves

Re: [patch] Error reporting during insert_breakpoints

[Pedro Alves]

On Wednesday 03 June 2009 22:46:44, Pedro Alves wrote:
> > @@ -1223,10 +1223,10 @@ Note: automatically using hardware break
> >               else
> >                 {
> >                   fprintf_unfiltered (tmp_error_stream, 
> > -                                     "Cannot insert breakpoint %d.\n", 
> > +                                     "Cannot insert breakpoint %d", 
> >                                       bpt->owner->number);
> >                   fprintf_filtered (tmp_error_stream, 
> > -                                   "Error accessing memory address ");
> > +                                   " at memory address ");
> >                   fputs_filtered (paddress (bpt->address), tmp_error_stream);
> >                   fprintf_filtered (tmp_error_stream, ": %s.\n",
> >                                     safe_strerror (val));
> 
> I'm not objecting to the change, but, honestly, I don't see where
> the assumption was, or where it is being removed?

Oh, silly me.  It's the s/accessing/at/.  Ignore me there.  BTW, since
you're touching this, how about merging all those fprintf calls
into a single call, and wrapping the user visible string in _() for
translation?

-- 
Pedro Alves

Re: [patch] Error reporting during insert_breakpoints

[Michael Snyder]

Pedro Alves wrote:
> On Wednesday 03 June 2009 21:42:14, Michael Snyder wrote:
>> Two small changes
>>
>> 1) If insert_breakpoints fails during stepping, "keep_going()"
>> contains a TRY_CATCH block, but does not print the exception error.
>> If the same thing happens during continue, "catch_errors" reports it.
> 
> Yes, please.  This silent stop bites me often, when longjmp-resume
> breakpoints fail to insert due to glibc pointer mangling...

Fell through the cracks, checking in now.