On 08.07.20 22:37, Eric Blake wrote:
> On 7/8/20 2:46 PM, Andrey Shinkevich wrote:
>>
>> On 08.07.2020 20:32, Eric Blake wrote:
>>> On 7/8/20 12:20 PM, Andrey Shinkevich wrote:
>>>> On 25.06.2020 18:21, Max Reitz wrote:
>>>>> v6:
>>>>> https://lists.nongnu.org/archive/html/qemu-devel/2019-08/msg01715.html
>>>>>
>>>>> Branch: https://github.com/XanClic/qemu.git child-access-functions-v7
>>>>> Branch: https://git.xanclic.moe/XanClic/qemu.git
>>>>> child-access-functions-v7
>>>>>
>>>>>
>>>> I cloned the branch from the github and built successfully.
>>>>
>>>> Running the iotests reports multiple errors of such a kind:
>>>>
>>>> 128: readarray -td '' formatting_line < <(sed -e 's/, fmt=/\x0/')
>>>>
>>>> "./common.filter: line 128: readarray: -d: invalid option"
>>>>
>>>
>>> Arrgh. If I'm reading bash's changelog correctly, readarray -d was
>>> introduced in bash 4.4, so I'm guessing you're still on 4.3 or
>>> earlier? What bash version and platform are you using?
>>>
>> My bash version is 4.2.46.
>>
>> It is the latest in the virtuozzolinux-base repository. I should
>> install the 4.4 package manually.
> 
> Well, if bash 4.2 is the default installed version on any of our
> platforms that meet our supported criteria, then we should instead fix
> the patch in question to avoid non-portable use of readarray.
> 
> Per https://repology.org/project/bash/versions (hinted from
> docs/system/build-platforms.rst), at least CentOS 7 still ships bash
> 4.2, and per 'make docker', centos7 is still a viable build target.  So
> we do indeed need to fix our regression.

There is no regression.  It’s just that I based this series on an
earlier version of “Make _filter_img_create more active” – when I sent a
pull request for that version, Peter already reported to me that it
failed on some test environments, so I revised it.

You’ll find there is no a7399eb in master; it’s 57ee95ed4ee there and
doesn’t use -d.

(My branch on github/gitea is still based on that older version, though,
because that’s what I wrote it on.)

Max