Re: [PATCH] tests: avoid FP failure when cp fails for /proc/cpuinfo
From:
Pádraig Brady
Subject:
Re: [PATCH] tests: avoid FP failure when cp fails for /proc/cpuinfo
Date:
Wed, 08 Jan 2014 23:56:07 +0000
User-agent:
Mozilla/5.0 (X11; Linux x86_64; rv:17.0) Gecko/20130110 Thunderbird/17.0.2
On 01/08/2014 10:48 PM, Bernhard Voelker wrote:
> On 01/08/2014 11:25 PM, Sergio Durigan Junior wrote:
>> On Wednesday, January 08 2014, Bernhard Voelker wrote:
>>> What does 'cp /proc/cpuinfo exp' produce?
>>
>> It produces "exp" with the contents of /proc/cpuinfo.
>
> Thanks again.
> Okay, together with the identical inode numbers you see
> that means that it is a problem of the aarch64 emulation
> on the OpenBuildService of SUSE rather than a real problem
> in cp. Good!
> Anyway, I think therefore skipping the test is better than
> a false positive.
>
> Have a nice day,
> Berny
Yes it seems that qemu-arm64 emulates access to /proc/cpuinfo:
https://github.com/susematz/qemu/blob/aarch64-1.6/linux-user/syscall.c
This passes for me on an aarch64 baremetal system:
test "$(stat -c %i /proc/cpuinfo)" = "$(stat -c %i /proc/cpuinfo)" ||
skip_ '(emulated?) /proc/cpuinfo inode number changed.'
Your cp error comparison is more direct and is better
protection against this issue.
So please push your patch.
thanks!
Pádraig.