[patch] Fix solib-display.exp crash

[patch] Fix solib-display.exp crash

[Jan Kratochvil]

Hi,

it is a testsuite regression against gdb-7.0 although the bug was probably
present even in gdb-7.0 (it started to reproduce since one completely
unrelated Daniel J.'s patch).  This patch fixes (-lmcheck may be required):

 PASS: gdb.base/solib-display.exp: display a_global
 PASS: gdb.base/solib-display.exp: display b_global
 PASS: gdb.base/solib-display.exp: display c_global
-ERROR: Process no longer exists
-UNRESOLVED: gdb.base/solib-display.exp: after rerun
-WARNING: remote_expect statement without a default case?!
-FAIL: gdb.base/solib-display.exp: Can't run to main (3)
+PASS: gdb.base/solib-display.exp: after rerun
+PASS: gdb.base/solib-display.exp: after rerun
+PASS: gdb.base/solib-display.exp: break 25
+PASS: gdb.base/solib-display.exp: continue
+PASS: gdb.base/solib-display.exp: display main_global
+PASS: gdb.base/solib-display.exp: display a_local
+PASS: gdb.base/solib-display.exp: display a_static
+PASS: gdb.base/solib-display.exp: break 25
+PASS: gdb.base/solib-display.exp: continue

No regressions on {x86_64,x86_64-m32,i686}-fedora12-linux-gnu.

The question is whether SYMBOL_SYMTAB (symbol) can be ever NULL at this point.
IMO not, it can be NULL only temporarily during reading symbols.


Thanks,
Jan


2010-01-20  Jan Kratochvil  <jan.kratochvil@redhat.com>

	* printcmd.c (display_uses_solib_p): Remove variable section.  Access
	objfile via SYMBOL_SYMTAB.

--- a/gdb/printcmd.c
+++ b/gdb/printcmd.c
@@ -1888,15 +1888,14 @@ display_uses_solib_p (const struct display *d,
 	{
 	  const struct block *const block = elts[i + 1].block;
 	  const struct symbol *const symbol = elts[i + 2].symbol;
-	  const struct obj_section *const section =
-	    SYMBOL_OBJ_SECTION (symbol);
 
 	  if (block != NULL
 	      && solib_contains_address_p (solib,
 					   block->startaddr))
 	    return 1;
 
-	  if (section && section->objfile == solib->objfile)
+	  /* SYMBOL_OBJ_SECTION (symbol) may be NULL.  */
+	  if (SYMBOL_SYMTAB (symbol)->objfile == solib->objfile)
 	    return 1;
 	}
       endpos -= oplen;

Re: [patch] Fix solib-display.exp crash

[Daniel Jacobowitz]

On Wed, Jan 20, 2010 at 12:40:01AM +0100, Jan Kratochvil wrote:
> The question is whether SYMBOL_SYMTAB (symbol) can be ever NULL at this point.
> IMO not, it can be NULL only temporarily during reading symbols.

I think that's right; this patch is fine.  Thanks!

-- 
Daniel Jacobowitz
CodeSourcery

[patch] Comment symbol->symtab NULL [Re: [patch] Fix solib-display.exp crash]

[Jan Kratochvil]

On Tue, 26 Jan 2010 22:24:25 +0100, Daniel Jacobowitz wrote:
> On Wed, Jan 20, 2010 at 12:40:01AM +0100, Jan Kratochvil wrote:
> > The question is whether SYMBOL_SYMTAB (symbol) can be ever NULL at this point.
> > IMO not, it can be NULL only temporarily during reading symbols.
> 
> I think that's right; this patch is fine.  Thanks!

Checked-in:
	http://sourceware.org/ml/gdb-cvs/2010-01/msg00217.html


OK to check-in this comment?


Thanks,
Jan

2010-01-27  Jan Kratochvil  <jan.kratochvil@redhat.com>

	* symtab.h (struct symbol <symtab>): New comment on NULL values.

--- a/gdb/symtab.h
+++ b/gdb/symtab.h
@@ -554,7 +554,8 @@ struct symbol
   struct type *type;
 
   /* The symbol table containing this symbol.  This is the file
-     associated with LINE.  */
+     associated with LINE.  It can be NULL during symbols read-in but it is
+     never NULL during normal operation.  */
   struct symtab *symtab;
 
   /* Domain code.  */

Re: [patch] Comment symbol->symtab NULL [Re: [patch] Fix solib-display.exp crash]

[Daniel Jacobowitz]

On Wed, Jan 27, 2010 at 12:50:16AM +0100, Jan Kratochvil wrote:
> On Tue, 26 Jan 2010 22:24:25 +0100, Daniel Jacobowitz wrote:
> > On Wed, Jan 20, 2010 at 12:40:01AM +0100, Jan Kratochvil wrote:
> > > The question is whether SYMBOL_SYMTAB (symbol) can be ever NULL at this point.
> > > IMO not, it can be NULL only temporarily during reading symbols.
> > 
> > I think that's right; this patch is fine.  Thanks!
> 
> Checked-in:
> 	http://sourceware.org/ml/gdb-cvs/2010-01/msg00217.html
> 
> 
> OK to check-in this comment?

Sure.

-- 
Daniel Jacobowitz
CodeSourcery

Re: [patch] Comment symbol->symtab NULL [Re: [patch] Fix solib-display.exp crash]

[Jan Kratochvil]

On Wed, 27 Jan 2010 01:05:55 +0100, Daniel Jacobowitz wrote:
> On Wed, Jan 27, 2010 at 12:50:16AM +0100, Jan Kratochvil wrote:
> > OK to check-in this comment?
> 
> Sure.

Checked-in:
	http://sourceware.org/ml/gdb-cvs/2010-01/msg00221.html


Thanks,
Jan