Test for GCC debug symbol bug

Test for GCC debug symbol bug

[law]

This patch adds a test for a GCC bug in its line number output.

Basically if we have a multi-line IF or WHILE conditional, GCC will
emit incorrect line numbers for the IF/WHILE statement.

The incorrect line numbers (of course) make debugging multi-line 
IF/WHILE statements a PITA.

To help ensure the GCC team doesn't break this again, I'd like to 
add this relatively simple test to the GDB testsuite.


	* gdb.base/break.c (multi_line_if_conditional): New function.
	(multi_ilne_while_conditional): Likewise.
	* gdb.base/break.exp: Verify that a breakpoint on a multi-line
	IF or WHILE condition puts the breakpoint at the start of
	the condition.

Index: break.c
===================================================================
RCS file: /cvs/cvsfiles/devo/gdb/testsuite/gdb.base/break.c,v
retrieving revision 1.7
diff -c -3 -p -r1.7 break.c
*** break.c	1999/10/02 00:24:35	1.7
--- break.c	2002/01/04 21:56:33
*************** int value;
*** 99,101 ****
--- 99,131 ----
      return (value);
  }
  
+ #ifdef PROTOTYPES
+ int multi_line_if_conditional (int a, int b, int c)
+ #else
+ int multi_line_if_conditional (a, b, c)
+   int a, b, c;
+ #endif
+ {
+   if (a
+       && b
+       && c)
+     return 0;
+   else
+     return 1;
+ }
+ 
+ #ifdef PROTOTYPES
+ int multi_line_while_conditional (int a, int b, int c)
+ #else
+ int multi_line_while_conditional (a, b, c)
+   int a, b, c;
+ #endif
+ {
+   while (a
+       && b
+       && c)
+     {
+       a--, b--, c--;
+     }
+   return 0;
+ }
Index: break.exp
===================================================================
RCS file: /cvs/cvsfiles/devo/gdb/testsuite/gdb.base/break.exp,v
retrieving revision 1.62.80.1
diff -c -3 -p -r1.62.80.1 break.exp
*** break.exp	2001/12/19 20:39:46	1.62.80.1
--- break.exp	2002/01/04 21:56:33
*************** gdb_test "break $srcfile:85" \
*** 122,127 ****
--- 122,139 ----
  
  
  #
+ # Test putting a break at the start of a multi-line if conditional.
+ # Verify the breakpoint was put at the start of the conditional.
+ #
+ gdb_test "break multi_line_if_conditional" \
+     "Breakpoint.*at.* file .*$srcfile, line 109\\." \
+     "breakpoint at start of multi line if conditional"
+ 
+ gdb_test "break multi_line_while_conditional" \
+     "Breakpoint.*at.* file .*$srcfile, line 124\\." \
+     "breakpoint at start of multi line while conditional"
+ 
+ #
  # check to see what breakpoints are set
  #
  if [target_info exists gdb_stub] {
*************** gdb_test "info break" \
*** 141,147 ****
  \[0-9\]+\[\t \]+breakpoint     keep y.* in factorial$proto at .*$srcfile:96.*
  \[0-9\]+\[\t \]+breakpoint     keep y.* in main at .*$srcfile:79.*
  \[0-9\]+\[\t \]+breakpoint     keep y.* in main at .*$srcfile:79.*
! \[0-9\]+\[\t \]+breakpoint     keep y.* in main at .*$srcfile:85" \
      "breakpoint info"
  
  
--- 153,161 ----
  \[0-9\]+\[\t \]+breakpoint     keep y.* in factorial$proto at .*$srcfile:96.*
  \[0-9\]+\[\t \]+breakpoint     keep y.* in main at .*$srcfile:79.*
  \[0-9\]+\[\t \]+breakpoint     keep y.* in main at .*$srcfile:79.*
! \[0-9\]+\[\t \]+breakpoint     keep y.* in main at .*$srcfile:85.*
! \[0-9\]+\[\t \]+breakpoint     keep y.* in multi_line_if_conditional at .*$srcfile:109.*
! \[0-9\]+\[\t \]+breakpoint     keep y.* in multi_line_while_conditional at .*$srcfile:124" \
      "breakpoint info"
  
  

Re: Test for GCC debug symbol bug

[Daniel Jacobowitz]

On Fri, Jan 04, 2002 at 03:14:59PM -0700, law@redhat.com wrote:
> 
> This patch adds a test for a GCC bug in its line number output.
> 
> Basically if we have a multi-line IF or WHILE conditional, GCC will
> emit incorrect line numbers for the IF/WHILE statement.
> 
> The incorrect line numbers (of course) make debugging multi-line 
> IF/WHILE statements a PITA.
> 
> To help ensure the GCC team doesn't break this again, I'd like to 
> add this relatively simple test to the GDB testsuite.
> 
> 
> 	* gdb.base/break.c (multi_line_if_conditional): New function.
> 	(multi_ilne_while_conditional): Likewise.
> 	* gdb.base/break.exp: Verify that a breakpoint on a multi-line
> 	IF or WHILE condition puts the breakpoint at the start of
> 	the condition.

<Andrew Cagney Hat On, apologies to Andrew>

This is obvious; please do commit it.  I'm all in favor of debug info
tests.

Of course, we may want to mark it XFAIL if we end up stopping where
broken GCC's would have; I really dislike FAILs that GDB can do nothing
about, although in some cases they're unavoidable.  Do we have a
documented policy for what all the DejaGNU result codes mean in the GDB
testsuite?  If not, should I propose one?

-- 
Daniel Jacobowitz                           Carnegie Mellon University
MontaVista Software                         Debian GNU/Linux Developer