diff options
| author | Matthew Heon <matthew.heon@pm.me> | 2020-07-29 14:39:02 -0400 |
|---|---|---|
| committer | Matthew Heon <mheon@redhat.com> | 2020-10-20 09:20:33 -0400 |
| commit | 3858fc1d019d13fe3378c129d133fd22789ac0dc (patch) | |
| tree | b4352cd8f7ee06586fcdfedd0e78a292ed644b54 /contrib/rootless-cni-infra | |
| parent | 338d521782727daca90efc2601b16502626aa53c (diff) | |
| download | podman-3858fc1d019d13fe3378c129d133fd22789ac0dc.tar.gz podman-3858fc1d019d13fe3378c129d133fd22789ac0dc.tar.bz2 podman-3858fc1d019d13fe3378c129d133fd22789ac0dc.zip | |
Use runtime names instead of paths in E2E tests
My patches to fix `--runtime /usr/bin/crun` being allowed to use
a different version of the crun runtime revealed a problem: we
were actually relying on that exact behavior in our E2E tests. We
specified the runtime path as `/usr/bin/runc` for the Ubuntu
tests, but that didn't exist, so Podman was actively looking for
a different, usable runc binary and using that, instead of the
path we explicitly hardcoded. Fixing the bug broke this, and thus
broke the tests.
Instead of hard-coding OCI runtime paths, swap to just using the
runtime name, `runc` or `crun`, and letting Podman figure out
where the runtime lives - it's quite good at that. This should
un-break the tests and make them more durable.
Signed-off-by: Matthew Heon <matthew.heon@pm.me>
Diffstat (limited to 'contrib/rootless-cni-infra')
0 files changed, 0 insertions, 0 deletions