Re: [PATCH v3 29/33] docker: remove 'deprecated' image definitions

[John Snow]

On 9/26/19 12:08 PM, Philippe Mathieu-Daudé wrote:
> On 9/26/19 6:04 PM, Alex Bennée wrote:
>>
>> Philippe Mathieu-Daudé <address@hidden> writes:
>>
>>> Hi Alex,
>>>
>>> On 9/26/19 1:34 AM, Alex Bennée wrote:
>>>> Philippe Mathieu-Daudé <address@hidden> writes:
>>>>> On 9/24/19 11:01 PM, Alex Bennée wrote:
>>>>>> From: John Snow <address@hidden>
>>>>>>
>>>>>> There isn't a debian.dockerfile anymore,
>>>>>> so perform some ghost-busting.
>>>>>
>>>>> Won't we deprecate other images in the future?
>>>>
>>>> Sure but we can just drop them from dockerfiles. It's not like we
>>>> allowed people to use them as we filtered them out.
>>>
>>> This patch isn't about removing a deprecated image, but about removing
>>> the handy DOCKER_DEPRECATED_IMAGES variable used to start a deprecation
>>> process.
>>
>> But it wasn't really doing anything. Because adding the image to
>> DOCKER_DEPRECATED_IMAGES had the same effect as using
>> DOCKER_PARTIAL_IMAGES. You couldn't invoke the test from make and it was
>> dropped out from the dependencies for the do-X-on-all image rules. As it
>> was also hidden from the help you might as well just delete the thing
>> all together.
>>
>>> Fam remembered once we should respect the QEMU deprecation policy even
>>> with docker images, because there might be users relying on them, so we
>>> want to give them time to adapt. I can not find a thread on the list, so
>>> we might have discussed that over IRC. The related commits are:
>>>
>>> $ git show bcaf457786c
>>>
>>>     docker: do not display deprecated images in 'make docker' help
>>>
>>>     the 'debian' base image is deprecated since 3e11974988d8
>>>
>>> $ git show 3e11974988d8
>>>
>>>     docker: warn users to use newer debian8/debian9 base image
>>>
>>>     to stay backward incompatible.
>>>
>>> I'd rather keep the DOCKER_DEPRECATED_IMAGES variable empty, maybe with
>>> a comment describing why it exists. What do you think?
>>
>> I think deprecations are better handled by a noisy warning in the
>> dockerfile than the silent suppression of their existence by the make
>> system.
> 
> Yes, fair enough. Thanks for keeping explaining ;)
> 
> Acked-by: Philippe Mathieu-Daudé <address@hidden>
> 
Also; I don't think testing utilities are beholden to the deprecation
process, so I don't think we need it around.

If we need to redirect our own developers in the future, I think we'll
be able to handle it on a case-by-case basis fairly well -- there's not
much framework that's getting dropped here.

Thanks, and sorry for causing so much fracas in the docker folder; but I
appreciate the chance to help make the tools a little easier to use and
for your help in deprecating Python2, which (in part) inspired this work.

--js