Re: one-too-many location in breakpoint

["Maciej W. Rozycki" <macro@codesourcery.com>]

On Tue, 29 Nov 2011, Tom Tromey wrote:

> Tom> I fixed all the other problems.  I'm running a regression test to make
> Tom> sure I didn't make something else break while doing this.  If all goes
> Tom> well I will send out refreshed patches.
> 
> I needed a couple of test suite changes to account for the different
> output.  Otherwise all is well.
> 
> This includes a test addition for the iterate_over_symbols change.
> 
> I didn't write a new test for the "normal_menu" problem.

 Just as a heads-up, this has regressed on the mips-sde-elf target (both 
MIPS32 and MIPS16 multilibs) as follows:

break A::outer::foo  (char*)  const if (a == 3)
No symbol "a" in current context.
(gdb) FAIL: gdb.cp/ovsrch.exp: break A::outer::foo  (char*)  const if (a == 3)
break 'A::outer::foo  (char*)  const' if (a == 3)
No symbol "a" in current context.
(gdb) FAIL: gdb.cp/ovsrch.exp: break 'A::outer::foo  (char*)  const' if (a == 3)
[...]
break A::outer::foo  (int)  const if (a == 3)
No symbol "a" in current context.
(gdb) FAIL: gdb.cp/ovsrch.exp: break A::outer::foo  (int)  const if (a == 3)
break 'A::outer::foo  (int)  const' if (a == 3)
No symbol "a" in current context.
(gdb) FAIL: gdb.cp/ovsrch.exp: break 'A::outer::foo  (int)  const' if (a == 3)
[...]
break A::B::inner::foo  (char*)  const if (a == 3)
No symbol "a" in current context.
(gdb) FAIL: gdb.cp/ovsrch.exp: break A::B::inner::foo  (char*)  const if (a == 3)
break 'A::B::inner::foo  (char*)  const' if (a == 3)
No symbol "a" in current context.
(gdb) FAIL: gdb.cp/ovsrch.exp: break 'A::B::inner::foo  (char*)  const' if (a == 3)
[...]
break A::B::inner::foo  (int)  const if (a == 3)
No symbol "a" in current context.
(gdb) FAIL: gdb.cp/ovsrch.exp: break A::B::inner::foo  (int)  const if (a == 3)
break 'A::B::inner::foo  (int)  const' if (a == 3)
No symbol "a" in current context.
(gdb) FAIL: gdb.cp/ovsrch.exp: break 'A::B::inner::foo  (int)  const' if (a == 3)

No regressions on mips-linux-gnu though.

 This is SDE code:

800011e8 <A::stop_here()>:
800011e8:	27bdfff8 	addiu	sp,sp,-8
800011ec:	afbe0004 	sw	s8,4(sp)
800011f0:	03a0f021 	move	s8,sp
800011f4:	03c0e821 	move	sp,s8
800011f8:	8fbe0004 	lw	s8,4(sp)
800011fc:	27bd0008 	addiu	sp,sp,8
80001200:	03e00008 	jr	ra
80001204:	00000000 	nop

(breakpoint at 800011f4), and this is Linux code:

00400790 <A::stop_here()>:
  400790:	27bdfff8 	addiu	sp,sp,-8
  400794:	afbe0004 	sw	s8,4(sp)
  400798:	03a0f021 	move	s8,sp
  40079c:	03c0e821 	move	sp,s8
  4007a0:	8fbe0004 	lw	s8,4(sp)
  4007a4:	27bd0008 	addiu	sp,sp,8
  4007a8:	03e00008 	jr	ra
  4007ac:	00000000 	nop

(breakpoint at 40079c).  And I fail to see "a" being defined anywhere 
throughout gdb.cp/ovsrch*, so I am curious as to why it succeeds for any 
target (including mips-linux-gnu) in the first place.

 I can see you made this adjustment:

-    gdb_test "break ${class}::foo if (a == 3)" "Breakpoint (\[0-9\]).*"
+    gdb_test "break ${class}::hibob if (a_param == 3)" "Breakpoint

but there's this definition elsewhere in this test case:

    set conditional1 "if (a == 3)"

as well, that you have left intact.  Should it be adjusted accordingly?  
Why is this condition accepted for some targets and not for some others 
anyway?

 Thoughts?

  Maciej