armhf testsuite failures possibly due to 7ad0ae7f42

armhf testsuite failures possibly due to 7ad0ae7f42

[Kyrill Tkachov]

Hi all,

Matthias' testing of libffi [1] (thanks!) shows some recent failures on 
armhf relating to closures with short types
(search for "FAIL:")

Considering that the results from May look okay I suspect the culprit is 
commit 7ad0ae7f42f3e208431ab66a9032dc9549f978d0 (arm: zext return value 
parameters)
from Oct 10.
Would it be possible to take a look and fix the failures?

Thanks,
Kyrill

(Please keep me on CC to any replies as I am not subscribed to this list)

[1] 
https://buildd.debian.org/status/fetch.php?pkg=libffi&arch=armhf&ver=3.3~20171020-1&stamp=1508524302&raw=0

[2] 
https://buildd.debian.org/status/fetch.php?pkg=libffi&arch=armhf&ver=3.3~20170512-1&stamp=1494644748&raw=0

Re: armhf testsuite failures possibly due to 7ad0ae7f42

[Saleem Abdulrasool]

On Mon, Oct 23, 2017 at 2:53 AM Kyrill Tkachov <kyrylo.tkachov@foss.arm.com>
wrote:

> Hi all,
>
> Matthias' testing of libffi [1] (thanks!) shows some recent failures on
> armhf relating to closures with short types
> (search for "FAIL:")
>
> Considering that the results from May look okay I suspect the culprit is
> commit 7ad0ae7f42f3e208431ab66a9032dc9549f978d0 (arm: zext return value
> parameters)
> from Oct 10.
> Would it be possible to take a look and fix the failures?
>

Thanks for the notification.

Yeah, the tests probably need to be adjusted.  The tests look like
something which would be impacted by the zext.  It was previously sending
whatever was in the register previously.


> Thanks,
> Kyrill
>
> (Please keep me on CC to any replies as I am not subscribed to this list)
>
> [1]
>
> https://buildd.debian.org/status/fetch.php?pkg=libffi&arch=armhf&ver=3.3~20171020-1&stamp=1508524302&raw=0
>
> [2]
>
> https://buildd.debian.org/status/fetch.php?pkg=libffi&arch=armhf&ver=3.3~20170512-1&stamp=1494644748&raw=0
>
-- 
Saleem Abdulrasool
compnerd (at) compnerd (dot) org

Re: armhf testsuite failures possibly due to 7ad0ae7f42

[Saleem Abdulrasool]

Hey, it was a silly bug caused by a level of indirection thing.  I sent a
PR (381) on github to address the issue.  It passes the tests locally.

On Mon, Oct 23, 2017 at 5:35 PM Saleem Abdulrasool <compnerd@compnerd.org>
wrote:

> On Mon, Oct 23, 2017 at 2:53 AM Kyrill Tkachov <
> kyrylo.tkachov@foss.arm.com> wrote:
>
>> Hi all,
>>
>> Matthias' testing of libffi [1] (thanks!) shows some recent failures on
>> armhf relating to closures with short types
>> (search for "FAIL:")
>>
>> Considering that the results from May look okay I suspect the culprit is
>> commit 7ad0ae7f42f3e208431ab66a9032dc9549f978d0 (arm: zext return value
>> parameters)
>> from Oct 10.
>> Would it be possible to take a look and fix the failures?
>>
>
> Thanks for the notification.
>
> Yeah, the tests probably need to be adjusted.  The tests look like
> something which would be impacted by the zext.  It was previously sending
> whatever was in the register previously.
>
>
>> Thanks,
>> Kyrill
>>
>> (Please keep me on CC to any replies as I am not subscribed to this list)
>>
>> [1]
>>
>> https://buildd.debian.org/status/fetch.php?pkg=libffi&arch=armhf&ver=3.3~20171020-1&stamp=1508524302&raw=0
>>
>> [2]
>>
>> https://buildd.debian.org/status/fetch.php?pkg=libffi&arch=armhf&ver=3.3~20170512-1&stamp=1494644748&raw=0
>>
> --
> Saleem Abdulrasool
> compnerd (at) compnerd (dot) org
>
-- 
Saleem Abdulrasool
compnerd (at) compnerd (dot) org

Re: armhf testsuite failures possibly due to 7ad0ae7f42

[Kyrill Tkachov]

On 25/10/17 01:20, Saleem Abdulrasool wrote:
> Hey, it was a silly bug caused by a level of indirection thing.  I 
> sent a PR (381) on github to address the issue.  It passes the tests 
> locally.
>

Thanks for looking into it!
Kyrill


> On Mon, Oct 23, 2017 at 5:35 PM Saleem Abdulrasool 
> <compnerd@compnerd.org <mailto:compnerd@compnerd.org>> wrote:
>
>     On Mon, Oct 23, 2017 at 2:53 AM Kyrill Tkachov
>     <kyrylo.tkachov@foss.arm.com <mailto:kyrylo.tkachov@foss.arm.com>>
>     wrote:
>
>         Hi all,
>
>         Matthias' testing of libffi [1] (thanks!) shows some recent
>         failures on
>         armhf relating to closures with short types
>         (search for "FAIL:")
>
>         Considering that the results from May look okay I suspect the
>         culprit is
>         commit 7ad0ae7f42f3e208431ab66a9032dc9549f978d0 (arm: zext
>         return value
>         parameters)
>         from Oct 10.
>         Would it be possible to take a look and fix the failures?
>
>
>     Thanks for the notification.
>
>     Yeah, the tests probably need to be adjusted.  The tests look like
>     something which would be impacted by the zext.  It was previously
>     sending whatever was in the register previously.
>
>
>         Thanks,
>         Kyrill
>
>         (Please keep me on CC to any replies as I am not subscribed to
>         this list)
>
>         [1]
>         https://buildd.debian.org/status/fetch.php?pkg=libffi&arch=armhf&ver=3.3~20171020-1&stamp=1508524302&raw=0
>         <https://buildd.debian.org/status/fetch.php?pkg=libffi&arch=armhf&ver=3.3%7E20171020-1&stamp=1508524302&raw=0>
>
>         [2]
>         https://buildd.debian.org/status/fetch.php?pkg=libffi&arch=armhf&ver=3.3~20170512-1&stamp=1494644748&raw=0
>         <https://buildd.debian.org/status/fetch.php?pkg=libffi&arch=armhf&ver=3.3%7E20170512-1&stamp=1494644748&raw=0>
>
>     -- 
>     Saleem Abdulrasool
>     compnerd (at) compnerd (dot) org
>
> -- 
> Saleem Abdulrasool
> compnerd (at) compnerd (dot) org