Re: [PATCH 3/3] docker: remove unused debian-sid and debian-ports

[John Snow]

On 9/20/19 5:19 PM, Alex Bennée wrote:
> 
> John Snow <address@hidden> writes:
> 
>> On 9/20/19 12:20 PM, Alex Bennée wrote:
>>>
>>> John Snow <address@hidden> writes:
>>>
>>>> On 9/20/19 4:49 AM, Philippe Mathieu-Daudé wrote:
>>>>> On 9/20/19 2:14 AM, John Snow wrote:
>>>>>> These are listed as "partial" images, but have no user.
>>>>>> Remove them.
>>>>>
>>>>> Well, I have WiP users from them. I could restore this content when they
>>>>> are ready... Ports is the base of deprecated Debian archs. On the other
>>>>> side Sid is the base for edge development I use from time to time to
>>>>> test latest gcc/binutils.
>>>>> I'll try to find time to raise WiP branches to PoC.
>>>>>
>>>>
>>>> I think that's the right thing to do. Right now, the docker tests
>>>> directory has a lot of stale entries and unusable tests. That might be
>>>> fine for the people working on it, but it makes it hard to understand
>>>> and use for those of us who only occasionally traipse into the directory.
>>>>
>>>> I'm removing all references to python2 -- but if there's no way for me
>>>> to test debian-sid and debian-ports, I can't test changes I need to make
>>>> to these "partial images", so they should be removed until they are
>>>> consumable.
>>>>
>>>> While I am sympathetic to the idea of having a library of partial images
>>>> to use for future tests, they're prone to rot if there's no way to
>>>> exercise them in-tree.
>>>
>>> Don't forget some "partial" images are only used for building TCG tests
>>> - we want to keep them. But as git is forever I can drop the sid/ports
>>> stuff for now until Phillipe has something to use them again.
>>>
>>
>> Oh, where? Is that defined within the tests/docker/Makefile.include or
>> in the TCG tree where I have not gone spelunking before?
> 
> Now they are selected by tests/tcg/configure.sh but generally they
> DISTRO-ARCH-cross images.
> 

I see -- "container_image" in configure.sh chooses them and consumes
them there.

Looks like that feeds to Makefile.qemu which defines
docker-build-guest-tests.

We call docker.py (DOCKER_SCRIPT) as:
docker.py cc ... -i qemu:$(DOCKER_IMAGE) ...

okay, so this bypasses the Docker Makefile pretty directly; but it looks
like it's only using images that the Makefile there doesn't count as
"Partial" universally.

You *do* have a section that marks images as "PARTIAL" if they're the
wrong architecture -- maybe it would be a semantic improvement to list
these as simply "DISABLED" or "UNAVAILABLE" -- the implication being
that they're complete images, but will not work for the current host
architecture.

Thanks for the pointer!

--js