[patch] Fix for failing gdb.base/mips_pro.exp under gcc-4.3.1

[patch] Fix for failing gdb.base/mips_pro.exp under gcc-4.3.1

[Paul Pluzhnikov]

Greetings,

Gcc-4.3.1 optimizes the whole thing away:

(gdb) disas main
Dump of assembler code for function main:
0x0000000000400340 <main+0>:    mov    $0x16,%eax
0x0000000000400345 <main+5>:    retq   
End of assembler dump.

Attached patch disables inlining.

The -O2 flag was added here:

  Sun Mar 17 13:35:31 1996  Fred Fish  <fnf@cygnus.com>

        * gdb.base/mips_pro.exp: Create mips_pro.ci to get gcc_compiled
        defined, and use it to compile the test case with -O2.  The
        native compilation still uses no optimization.

I could not find *why* optimization is necessary for that test case,
so don't know if disabling inlining is the correct fix :(

--
Paul Pluzhnikov

2008-02-12  Paul Pluzhnikov  <ppluzhnikov@google.com>

	*gdb.base/mips_pro.exp: compile with gcc -fno-inline,
	lest gcc-4.3.1 optimizes the whole thing away.
	

Index: gdb.base/mips_pro.exp
===================================================================
RCS file: /cvs/src/src/gdb/testsuite/gdb.base/mips_pro.exp,v
retrieving revision 1.10
diff -u -p -u -r1.10 mips_pro.exp
--- gdb.base/mips_pro.exp	6 Aug 2008 12:52:07 -0000	1.10
+++ gdb.base/mips_pro.exp	22 Aug 2008 00:20:15 -0000
@@ -35,7 +35,8 @@ if [get_compiler_info ${binfile}] {
 # This test must be compiled with -O2 if using gcc.
 
 if { [test_compiler_info gcc-*-*] } then {
-    if  { [gdb_compile "${srcdir}/$subdir/${srcfile}" "${binfile}" executable {debug additional_flags=-O2}] != "" } {
+    if  { [gdb_compile "${srcdir}/$subdir/${srcfile}" "${binfile}" \
+	    executable {debug additional_flags=-O2 additional_flags=-fno-inline}] != "" } {
      untested mips_pro.exp
      return -1
     }

Re: [patch] Fix for failing gdb.base/mips_pro.exp under gcc-4.3.1

[Michael Snyder]

Paul Pluzhnikov wrote:
> Greetings,
> 
> Gcc-4.3.1 optimizes the whole thing away:
> 
> (gdb) disas main
> Dump of assembler code for function main:
> 0x0000000000400340 <main+0>:    mov    $0x16,%eax
> 0x0000000000400345 <main+5>:    retq
> End of assembler dump.
> 
> Attached patch disables inlining.
> 
> The -O2 flag was added here:
> 
>   Sun Mar 17 13:35:31 1996  Fred Fish  <fnf@cygnus.com>
> 
>         * gdb.base/mips_pro.exp: Create mips_pro.ci to get gcc_compiled
>         defined, and use it to compile the test case with -O2.  The
>         native compilation still uses no optimization.
> 
> I could not find *why* optimization is necessary for that test case,
> so don't know if disabling inlining is the correct fix :(

Well, I think it's supposed to be specifically
a test for debugging optimized, no-symbols code.

But inlining would probably be a problem, yes...


> 
> --
> Paul Pluzhnikov
> 
> 2008-02-12  Paul Pluzhnikov  <ppluzhnikov@google.com>
> 
>         *gdb.base/mips_pro.exp: compile with gcc -fno-inline,
>         lest gcc-4.3.1 optimizes the whole thing away.
> 
> 
> Index: gdb.base/mips_pro.exp
> ===================================================================
> RCS file: /cvs/src/src/gdb/testsuite/gdb.base/mips_pro.exp,v
> retrieving revision 1.10
> diff -u -p -u -r1.10 mips_pro.exp
> --- gdb.base/mips_pro.exp       6 Aug 2008 12:52:07 -0000       1.10
> +++ gdb.base/mips_pro.exp       22 Aug 2008 00:20:15 -0000
> @@ -35,7 +35,8 @@ if [get_compiler_info ${binfile}] {
>  # This test must be compiled with -O2 if using gcc.
> 
>  if { [test_compiler_info gcc-*-*] } then {
> -    if  { [gdb_compile "${srcdir}/$subdir/${srcfile}" "${binfile}" executable {debug additional_flags=-O2}] != "" } {
> +    if  { [gdb_compile "${srcdir}/$subdir/${srcfile}" "${binfile}" \
> +           executable {debug additional_flags=-O2 additional_flags=-fno-inline}] != "" } {
>       untested mips_pro.exp
>       return -1
>      }

Re: [patch] Fix for failing gdb.base/mips_pro.exp under gcc-4.3.1

[Paul Pluzhnikov]

Ping,

http://sourceware.org/ml/gdb-patches/2008-08/msg00576.html

-- 
Paul Pluzhnikov

Re: [patch] Fix for failing gdb.base/mips_pro.exp under gcc-4.3.1

[Joel Brobecker]

(sorry about the delay - it looks like everyone was very very busy
this summer)

> I could not find *why* optimization is necessary for that test case,
> so don't know if disabling inlining is the correct fix :(

Like Michael said, we have to assume the obvious, which is that they
wanted to test unwinding of optimized code.

> 2008-02-12  Paul Pluzhnikov  <ppluzhnikov@google.com>
> 
> 	*gdb.base/mips_pro.exp: compile with gcc -fno-inline,
> 	lest gcc-4.3.1 optimizes the whole thing away.

With that in mind, your patch makes sense. We want optimized code, but
not that much optimization :).

-- 
Joel

Re: [patch] Fix for failing gdb.base/mips_pro.exp under gcc-4.3.1

[Paul Pluzhnikov]

On Thu, Sep 4, 2008 at 4:58 PM, Joel Brobecker <brobecker@adacore.com> wrote:
> (sorry about the delay - it looks like everyone was very very busy
> this summer)
>
>> I could not find *why* optimization is necessary for that test case,
>> so don't know if disabling inlining is the correct fix :(
>
> Like Michael said, we have to assume the obvious, which is that they
> wanted to test unwinding of optimized code.
>
>> 2008-02-12  Paul Pluzhnikov  <ppluzhnikov@google.com>

Hmm, that should have been 2008-08-22 ...

>>
>>       *gdb.base/mips_pro.exp: compile with gcc -fno-inline,
>>       lest gcc-4.3.1 optimizes the whole thing away.
>
> With that in mind, your patch makes sense. We want optimized code, but
> not that much optimization :).

So ok to commit?

Thanks,
-- 
Paul Pluzhnikov

Re: [patch] Fix for failing gdb.base/mips_pro.exp under gcc-4.3.1

[Joel Brobecker]

> >>       *gdb.base/mips_pro.exp: compile with gcc -fno-inline,
> >>       lest gcc-4.3.1 optimizes the whole thing away.
> >
> > With that in mind, your patch makes sense. We want optimized code, but
> > not that much optimization :).
> 
> So ok to commit?

Yes :).

-- 
Joel