Commit: Skip cfi-i386-d gas test for PE based targets

Message ID 87y2oxnc2y.fsf@redhat.com
State New
Headers show
Series
  • Commit: Skip cfi-i386-d gas test for PE based targets
Related show

Commit Message

Alan Modra via Binutils June 8, 2020, 9:25 a.m.
Hi Guys,

  I am applying the patch below to a skip of the new cfi-i386-2 gas test
  for PE based targets, as these do not support the .type pseudo-op.
  (Well not as it is being used in the test).

Cheers
  Nick

gas/ChangeLog
2020-06-08  Nick Clifton  <nickc@redhat.com>

	* testsuite/gas/cfi/cfi-i386-2.d: Skip for PE based targets.

Comments

Jan Beulich June 8, 2020, 10:05 a.m. | #1
On 08.06.2020 11:25, Nick Clifton via Binutils wrote:
>   I am applying the patch below to a skip of the new cfi-i386-2 gas test

>   for PE based targets, as these do not support the .type pseudo-op.

>   (Well not as it is being used in the test).


Oh, I'm sorry for not having noticed / checked. I don't think the .type
directive is in any way meaningful for the test, so an alternative
might be to drop it.

Jan
Alan Modra via Binutils June 8, 2020, 10:10 a.m. | #2
Hi Jan,

> Oh, I'm sorry for not having noticed / checked. I don't think the .type

> directive is in any way meaningful for the test, so an alternative

> might be to drop it.


Ah - I did wonder about that.  But I was not sure if .cfi_startproc needed
a function symbol or not.  If it does not then please feel free to revert
my patch and just delete the .type directive instead.

Cheers
  Nick
Jan Beulich June 25, 2020, 1:55 p.m. | #3
On 08.06.2020 12:10, Nick Clifton wrote:
>> Oh, I'm sorry for not having noticed / checked. I don't think the .type

>> directive is in any way meaningful for the test, so an alternative

>> might be to drop it.

> 

> Ah - I did wonder about that.  But I was not sure if .cfi_startproc needed

> a function symbol or not.  If it does not then please feel free to revert

> my patch and just delete the .type directive instead.


Actually when making a patch for this I realized this 2nd test
uses readelf, not objdump, and at a guess that's why it was
introduced. So I'll leave this alone.

Jan

Patch

diff --git a/gas/testsuite/gas/cfi/cfi-i386-2.d b/gas/testsuite/gas/cfi/cfi-i386-2.d
index 05e414d93e..0e7f2533eb 100644
--- a/gas/testsuite/gas/cfi/cfi-i386-2.d
+++ b/gas/testsuite/gas/cfi/cfi-i386-2.d
@@ -1,5 +1,8 @@ 
 #readelf: -wf
 #name: CFI on i386, 2
+# PE based targets do not support the .type pseudo-op
+#notarget: *-*-mingw* *-*-cygwin* *-*-pe
+
 Contents of the .eh_frame section:
 
 00000000 0+0014 0+0000 CIE