[PATCH]: Fix LTO testsuite case on sparc

[PATCH]: Fix LTO testsuite case on sparc

[David Miller]

Like other tests that want to link against the __sync_* primitives, we
have to force "-mcpu=v9" on sparc to make sure those symbols will
actually be there.

Ok to commit?

2010-03-12  David S. Miller  <davem@davemloft.net>

	* gcc.dg/lto/20090313_0.c: Add -mcpu=v9 to dg-lto-options on
	sparc.

diff --git a/gcc/testsuite/gcc.dg/lto/20090313_0.c b/gcc/testsuite/gcc.dg/lto/20090313_0.c
index 120ba3c..523ae2d 100644
--- a/gcc/testsuite/gcc.dg/lto/20090313_0.c
+++ b/gcc/testsuite/gcc.dg/lto/20090313_0.c
@@ -1,3 +1,4 @@
+/* { dg-lto-options "-mcpu=v9" { target sparc*-*-* } } */
 void
 _cairo_clip_path_reference () {
   int a;

Re: [PATCH]: Fix LTO testsuite case on sparc

[Janis Johnson]

On Fri, Mar 12, 2010 at 3:47 PM, David Miller <davem@davemloft.net> wrote:
>
> Like other tests that want to link against the __sync_* primitives, we
> have to force "-mcpu=v9" on sparc to make sure those symbols will
> actually be there.
>
> Ok to commit?

Sure.  As a sparc maintainer you shouldn't need to ask about sparc-specific
changes.

Janis

> 2010-03-12  David S. Miller  <davem@davemloft.net>
>
>        * gcc.dg/lto/20090313_0.c: Add -mcpu=v9 to dg-lto-options on
>        sparc.
>
> diff --git a/gcc/testsuite/gcc.dg/lto/20090313_0.c b/gcc/testsuite/gcc.dg/lto/20090313_0.c
> index 120ba3c..523ae2d 100644
> --- a/gcc/testsuite/gcc.dg/lto/20090313_0.c
> +++ b/gcc/testsuite/gcc.dg/lto/20090313_0.c
> @@ -1,3 +1,4 @@
> +/* { dg-lto-options "-mcpu=v9" { target sparc*-*-* } } */
>  void
>  _cairo_clip_path_reference () {
>   int a;
>

Re: [PATCH]: Fix LTO testsuite case on sparc

[David Miller]

From: Janis Johnson <janis.marie.johnson@gmail.com>
Date: Fri, 12 Mar 2010 16:59:02 -0800

> On Fri, Mar 12, 2010 at 3:47 PM, David Miller <davem@davemloft.net> wrote:
>>
>> Like other tests that want to link against the __sync_* primitives, we
>> have to force "-mcpu=v9" on sparc to make sure those symbols will
>> actually be there.
>>
>> Ok to commit?
> 
> Sure.  As a sparc maintainer you shouldn't need to ask about sparc-specific
> changes.

Even so, it's nice to get a smoke test review from at least one other
person :-)

Thanks!

Re: [PATCH]: Fix LTO testsuite case on sparc

[Jakub Jelinek]

On Fri, Mar 12, 2010 at 03:47:29PM -0800, David Miller wrote:
> 
> Like other tests that want to link against the __sync_* primitives, we
> have to force "-mcpu=v9" on sparc to make sure those symbols will
> actually be there.

This should be instead
/* { dg-require-effective-target sync_int_long } */
IMNSHO.

> 2010-03-12  David S. Miller  <davem@davemloft.net>
> 
> 	* gcc.dg/lto/20090313_0.c: Add -mcpu=v9 to dg-lto-options on
> 	sparc.
> 
> diff --git a/gcc/testsuite/gcc.dg/lto/20090313_0.c b/gcc/testsuite/gcc.dg/lto/20090313_0.c
> index 120ba3c..523ae2d 100644
> --- a/gcc/testsuite/gcc.dg/lto/20090313_0.c
> +++ b/gcc/testsuite/gcc.dg/lto/20090313_0.c
> @@ -1,3 +1,4 @@
> +/* { dg-lto-options "-mcpu=v9" { target sparc*-*-* } } */
>  void
>  _cairo_clip_path_reference () {
>    int a;

	Jakub

Re: [PATCH]: Fix LTO testsuite case on sparc

[David Miller]

From: Jakub Jelinek <jakub@redhat.com>
Date: Sat, 13 Mar 2010 08:41:38 +0100

> On Fri, Mar 12, 2010 at 03:47:29PM -0800, David Miller wrote:
>> 
>> Like other tests that want to link against the __sync_* primitives, we
>> have to force "-mcpu=v9" on sparc to make sure those symbols will
>> actually be there.
> 
> This should be instead
> /* { dg-require-effective-target sync_int_long } */
> IMNSHO.

That won't work.

The LTO test framework only supports a tiny subset of
the usual dg-* expressions, not all of them.

lto.exp explicitly checks the options specified and only
accepts ones that it knows about.

I checked all of this out before writing my patch Jakub.
:-)