Re: [PATCH v2] tests/docker: ensure container command is probed at most once

["Daniel P. Berrangé" <berrange@redhat.com>]

On Wed, Jul 15, 2026 at 09:22:05AM -0700, Pierrick Bouvier wrote:
> On 7/15/2026 9:16 AM, Daniel P. Berrangé wrote:
> > On Wed, Jul 15, 2026 at 09:07:18AM -0700, Pierrick Bouvier wrote:
> >> On 7/15/2026 3:42 AM, Daniel P. Berrangé wrote:
> >>> The '--engine' arg accepts either 'podman' or 'docker', which is
> >>> not sufficiently granular to map directly to a command. This
> >>> means that  docker.py still has to then probe the exact command
> >>> to use.
> >>>
> >>> Meanwhile the 'probe' command prints out the full command to use
> >>> but this cannot be passed back to docker.py to avoid probing
> >>> again, so the caching is only useful in the few case where we
> >>> run a container directly bypassing docker.py.
> >>>
> >>> Address this by replacing --engine with --command for docker.py.
> >>>
> >>> This in turn requires the --container-engine configure arg to be
> >>> replaced with --container-command.
> >>>
> >>> With these changes the container command is probed at most once
> >>> during configure and never again, while running in an unconfigured
> >>> tree will still probe on demand.
> >>>
> >>> Signed-off-by: Daniel P. Berrangé <berrange@redhat.com>
> >>> ---
> >>>
> >>> In v2:
> >>>
> >>>  * Don't set container_command to "no" if 'probe' finds
> >>>    not viable binaries.
> >>>
> >>>  configure                     | 25 +++++++------
> >>>  tests/docker/Makefile.include | 13 ++++---
> >>>  tests/docker/docker.py        | 67 ++++++++++++-----------------------
> >>>  3 files changed, 40 insertions(+), 65 deletions(-)
> > 
> > 
> >>
> >> I still observe the same error when running make-tcg without
> >> podman/docker installed.
> >> bash: line 1: no: command not found
> >>
> >> And I will suggest again to include in your series:
> >> https://lore.kernel.org/qemu-devel/20260710214343.2065491-27-pierrick.bouvier@oss.qualcomm.com/
> >> and detect if probe command worked or failed, and adapt behavior based
> >> on it. Existing behavior to compare to a string "no" is a very
> >> unfamiliar pattern for any command. `which` does not return no for instance.
> > 
> > That didn't do anything for me when I tested it, and this current
> > patch did fix the problem. Are you testing on top of a vanilla
> > git master, or did you have other patches that aren't merged
> > upstream.
> >
> 
> Originally caught on GitHub.
> If you can survive to have to log on this evil platform:
> https://github.com/p-b-o/qemu-ci/actions/runs/29410261308/job/87335442476
> 
> Which I reproduced on my machine locally using the command I posted at
> the end of my email.
> Top of master + this patch, without any other patch.
> 
> > When /usr/bin/podman was absent
> > 
> >  $ ./configure --target-list=s390x-linux-user
> >  ...
> >  $ grep CONTAINER build/config-host.mak 
> >  <no match>
> >  
> >  $ make run-tcg-tests-s390x-linux-user
> >   changing dir to build for make "run-tcg-tests-s390x-linux-user"...
> >   make[1]: Entering directory '/home/berrange/src/virt/qemu/build'
> >   ninja: no work to do.
> >   /home/berrange/src/virt/qemu/build/pyvenv/bin/meson introspect --targets --tests --benchmarks | /home/berrange/src/virt/qemu/build/pyvenv/bin/python3 -B scripts/mtest2make.py > Makefile.mtest
> >   make[1]: *** No rule to make target 'run-tcg-tests-s390x-linux-user'.  Stop.
> > 
> > configure has correctly skipped setup of the tcg cross
> > compiler targets.
> > 
> > When /usr/bin/podman was re-instaled it once again
> > worked
> > 
> > 
> >   $ ./configure --target-list=s390x-linux-user
> >   ...
> >   $ grep CONTAINER build/config-host.mak 
> >   CONTAINER_COMMAND=podman
> > 
> >   $ make run-tcg-tests-s390x-linux-user
> >   changing dir to build for make "run-tcg-tests-s390x-linux-user"...
> >   make[1]: Entering directory '/home/berrange/src/virt/qemu/build'
> >   ninja: no work to do.
> >   /home/berrange/src/virt/qemu/build/pyvenv/bin/meson introspect --targets --tests --benchmarks | /home/berrange/src/virt/qemu/build/pyvenv/bin/python3 -B scripts/mtest2make.py > Makefile.mtest
> >     BUILD   debian-s390x-cross
> >     BUILD   s390x-linux-user guest-tests
> >   tests/tcg/s390x-linux-user: -march=z14 detected
> >   tests/tcg/s390x-linux-user: -march=z15 detected
> >   ...tests running...
> > 
> >> You can test next patch using podman (which it seems you have):
> >> ```
> >> podman build \
> >> -f tests/docker/dockerfiles/debian-all-test-cross.docker -t all
> >> podman run -it --rm all -w $(pwd) -v $(pwd):$(pwd) all \
> >> bash -c "./configure && cd build && make all -j32 &&make check-tcg -j32"
> >> ```
> >
> 
> Try to run the exact command above, and we'll see if you can reproduce
> or not.

Oh it is the bash-ism  test "==" vs "="

With regards,
Daniel
-- 
|: https://berrange.com       ~~        https://hachyderm.io/@berrange :|
|: https://libvirt.org          ~~          https://entangle-photo.org :|
|: https://pixelfed.art/berrange   ~~    https://fstop138.berrange.com :|