[PR,testsuite/91842] Skip gcc.dg/ipa/ipa-sra-19.c on power

Message ID ri6eezv1916.fsf@suse.cz
State New
Headers show
Series
  • [PR,testsuite/91842] Skip gcc.dg/ipa/ipa-sra-19.c on power
Related show

Commit Message

Martin Jambor Oct. 2, 2019, 3:06 p.m.
Hi,

I seem to remember I minimized gcc.dg/ipa/ipa-sra-19.c on power but
perhaps I am wrong because the testcase fails there with a
power-specific error:

gcc.dg/ipa/ipa-sra-19.c:19:3: error: AltiVec argument passed to unprototyped function

I am going to simply skip it there with the following patch, which I
hope is obvious.  Tested by running ipa.exp on both ppc64le-linux and
x86_64-linux.

Thanks,

Martin

Comments

Rainer Orth Oct. 2, 2019, 4:11 p.m. | #1
Hi Martin,

> I seem to remember I minimized gcc.dg/ipa/ipa-sra-19.c on power but

> perhaps I am wrong because the testcase fails there with a

> power-specific error:

>

> gcc.dg/ipa/ipa-sra-19.c:19:3: error: AltiVec argument passed to

> unprototyped function

>

> I am going to simply skip it there with the following patch, which I

> hope is obvious.  Tested by running ipa.exp on both ppc64le-linux and

> x86_64-linux.

[...]
> diff --git a/gcc/testsuite/gcc.dg/ipa/ipa-sra-19.c

> b/gcc/testsuite/gcc.dg/ipa/ipa-sra-19.c

> index adebaa5f5e1..d219411d8ba 100644

> --- a/gcc/testsuite/gcc.dg/ipa/ipa-sra-19.c

> +++ b/gcc/testsuite/gcc.dg/ipa/ipa-sra-19.c

> @@ -1,5 +1,6 @@

>  /* { dg-do compile } */

>  /* { dg-options "-O2"  } */

> +/* { dg-skip-if "" { powerpc*-*-* } } */


please use a short comment (first arg to dg-skip-if) explaining why you
skip the test, otherwise this will be anything but obvious shortly.

I'll leave approval to the PowerPC maintainers.

	Rainer

-- 
-----------------------------------------------------------------------------
Rainer Orth, Center for Biotechnology, Bielefeld University
Andreas Krebbel Oct. 24, 2019, 11:43 a.m. | #2
On 02.10.19 17:06, Martin Jambor wrote:
> Hi,

> 

> I seem to remember I minimized gcc.dg/ipa/ipa-sra-19.c on power but

> perhaps I am wrong because the testcase fails there with a

> power-specific error:

> 

> gcc.dg/ipa/ipa-sra-19.c:19:3: error: AltiVec argument passed to unprototyped function

> 

> I am going to simply skip it there with the following patch, which I

> hope is obvious.  Tested by running ipa.exp on both ppc64le-linux and

> x86_64-linux.

> 

> Thanks,

> 

> Martin

> 

> 

> diff --git a/gcc/testsuite/gcc.dg/ipa/ipa-sra-19.c b/gcc/testsuite/gcc.dg/ipa/ipa-sra-19.c

> index adebaa5f5e1..d219411d8ba 100644

> --- a/gcc/testsuite/gcc.dg/ipa/ipa-sra-19.c

> +++ b/gcc/testsuite/gcc.dg/ipa/ipa-sra-19.c

> @@ -1,5 +1,6 @@

>  /* { dg-do compile } */

>  /* { dg-options "-O2"  } */

> +/* { dg-skip-if "" { powerpc*-*-* } } */

>  

>  typedef int __attribute__((__vector_size__(16))) vectype;

> 

> 


I ran into the same problem on IBM Z. Is it important for the testcase to leave the argument list of
k unspecified or would it be ok to add it?

diff --git a/gcc/testsuite/gcc.dg/ipa/ipa-sra-19.c b/gcc/testsuite/gcc.dg/ipa/ipa-sra-19.c
index d219411d8ba..d9dcd33cb76 100644
--- a/gcc/testsuite/gcc.dg/ipa/ipa-sra-19.c
+++ b/gcc/testsuite/gcc.dg/ipa/ipa-sra-19.c
@@ -5,7 +5,7 @@
 typedef int __attribute__((__vector_size__(16))) vectype;

 vectype dk();
-vectype k();
+vectype k(vectype);

 int b;
 vectype *j;
Martin Jambor Oct. 24, 2019, 1:26 p.m. | #3
Hi,

On Thu, Oct 24 2019, Andreas Krebbel wrote:
> On 02.10.19 17:06, Martin Jambor wrote:

>> Hi,

>> 

>> I seem to remember I minimized gcc.dg/ipa/ipa-sra-19.c on power but

>> perhaps I am wrong because the testcase fails there with a

>> power-specific error:

>> 

>> gcc.dg/ipa/ipa-sra-19.c:19:3: error: AltiVec argument passed to unprototyped function

>> 

>> I am going to simply skip it there with the following patch, which I

>> hope is obvious.  Tested by running ipa.exp on both ppc64le-linux and

>> x86_64-linux.

>> 

>> Thanks,

>> 

>> Martin

>> 

>> 

>> diff --git a/gcc/testsuite/gcc.dg/ipa/ipa-sra-19.c b/gcc/testsuite/gcc.dg/ipa/ipa-sra-19.c

>> index adebaa5f5e1..d219411d8ba 100644

>> --- a/gcc/testsuite/gcc.dg/ipa/ipa-sra-19.c

>> +++ b/gcc/testsuite/gcc.dg/ipa/ipa-sra-19.c

>> @@ -1,5 +1,6 @@

>>  /* { dg-do compile } */

>>  /* { dg-options "-O2"  } */

>> +/* { dg-skip-if "" { powerpc*-*-* } } */

>>  

>>  typedef int __attribute__((__vector_size__(16))) vectype;

>> 

>> 

>

> I ran into the same problem on IBM Z. Is it important for the testcase to leave the argument list of

> k unspecified or would it be ok to add it?


I wanted to write to you that the un-prototypedness is on purpose and
essential to test what the bug was in the past but this time I actually
managed to find the associated fix in my ipa-sra branch and found out
that I mis-remembered, that it is not the case.  Sorry for not doing
that before.  I believe the patch is OK then and we can even remove the
dg-skip-if I added.  And by that I mean that although I'm not a
reviewer, I would consider it obvious.  Will you do it or should I take
care of it?

Thanks,

Martin


>

> diff --git a/gcc/testsuite/gcc.dg/ipa/ipa-sra-19.c b/gcc/testsuite/gcc.dg/ipa/ipa-sra-19.c

> index d219411d8ba..d9dcd33cb76 100644

> --- a/gcc/testsuite/gcc.dg/ipa/ipa-sra-19.c

> +++ b/gcc/testsuite/gcc.dg/ipa/ipa-sra-19.c

> @@ -5,7 +5,7 @@

>  typedef int __attribute__((__vector_size__(16))) vectype;

>

>  vectype dk();

> -vectype k();

> +vectype k(vectype);

>

>  int b;

>  vectype *j;
Andreas Krebbel Oct. 24, 2019, 3:17 p.m. | #4
On 24.10.19 15:26, Martin Jambor wrote:
> Hi,

> 

> On Thu, Oct 24 2019, Andreas Krebbel wrote:

>> On 02.10.19 17:06, Martin Jambor wrote:

>>> Hi,

>>>

>>> I seem to remember I minimized gcc.dg/ipa/ipa-sra-19.c on power but

>>> perhaps I am wrong because the testcase fails there with a

>>> power-specific error:

>>>

>>> gcc.dg/ipa/ipa-sra-19.c:19:3: error: AltiVec argument passed to unprototyped function

>>>

>>> I am going to simply skip it there with the following patch, which I

>>> hope is obvious.  Tested by running ipa.exp on both ppc64le-linux and

>>> x86_64-linux.

>>>

>>> Thanks,

>>>

>>> Martin

>>>

>>>

>>> diff --git a/gcc/testsuite/gcc.dg/ipa/ipa-sra-19.c b/gcc/testsuite/gcc.dg/ipa/ipa-sra-19.c

>>> index adebaa5f5e1..d219411d8ba 100644

>>> --- a/gcc/testsuite/gcc.dg/ipa/ipa-sra-19.c

>>> +++ b/gcc/testsuite/gcc.dg/ipa/ipa-sra-19.c

>>> @@ -1,5 +1,6 @@

>>>  /* { dg-do compile } */

>>>  /* { dg-options "-O2"  } */

>>> +/* { dg-skip-if "" { powerpc*-*-* } } */

>>>  

>>>  typedef int __attribute__((__vector_size__(16))) vectype;

>>>

>>>

>>

>> I ran into the same problem on IBM Z. Is it important for the testcase to leave the argument list of

>> k unspecified or would it be ok to add it?

> 

> I wanted to write to you that the un-prototypedness is on purpose and

> essential to test what the bug was in the past but this time I actually

> managed to find the associated fix in my ipa-sra branch and found out

> that I mis-remembered, that it is not the case.  Sorry for not doing

> that before.  I believe the patch is OK then and we can even remove the

> dg-skip-if I added.  And by that I mean that although I'm not a

> reviewer, I would consider it obvious.  Will you do it or should I take

> care of it?


I will do it. Thanks!

Andreas

> 

> Thanks,

> 

> Martin

> 

> 

>>

>> diff --git a/gcc/testsuite/gcc.dg/ipa/ipa-sra-19.c b/gcc/testsuite/gcc.dg/ipa/ipa-sra-19.c

>> index d219411d8ba..d9dcd33cb76 100644

>> --- a/gcc/testsuite/gcc.dg/ipa/ipa-sra-19.c

>> +++ b/gcc/testsuite/gcc.dg/ipa/ipa-sra-19.c

>> @@ -5,7 +5,7 @@

>>  typedef int __attribute__((__vector_size__(16))) vectype;

>>

>>  vectype dk();

>> -vectype k();

>> +vectype k(vectype);

>>

>>  int b;

>>  vectype *j;

Patch

diff --git a/gcc/testsuite/gcc.dg/ipa/ipa-sra-19.c b/gcc/testsuite/gcc.dg/ipa/ipa-sra-19.c
index adebaa5f5e1..d219411d8ba 100644
--- a/gcc/testsuite/gcc.dg/ipa/ipa-sra-19.c
+++ b/gcc/testsuite/gcc.dg/ipa/ipa-sra-19.c
@@ -1,5 +1,6 @@ 
 /* { dg-do compile } */
 /* { dg-options "-O2"  } */
+/* { dg-skip-if "" { powerpc*-*-* } } */
 
 typedef int __attribute__((__vector_size__(16))) vectype;