CI and system tests currently pull some images from docker.io.
Eliminate that, by:
- building a custom image containing much of what we need
for testing; and
- copying other needed images to quay.io
(Reason: effective 2020-11-01 docker.io will limit the
number of image pulls).
The principal change is to create a new quay.io/libpod/testimage,
using the new test/system/build-testimage script, instead of
relying on quay.io/libpod/alpine_labels. We also switch to
using a hardcoded :YYYYMMDD tag, instead of :latest, in an
attempt to futureproof our CI. This image includes 'httpd'
from busybox-extras, which we use in our networking test
(previously we had to pull and run busybox from docker.io).
The testimage can and should be extended as needed for future
tests, e.g. adding test file content or other useful tools.
For the '--pull' tests which require actually pulling from
the registry, I've created an image with the same name but
tagged :00000000 so it will never be pulled by default.
Since this image is only used minimally, it's just busybox.
Unfortunately there remain two cases we cannot solve in
this tiny alpine-based image:
1) docker registry
2) systemd
For those, I've (manually) run:
podman pull [ docker.io/library/registry:2.7 | registry.fedoraproject.org/fedora:31 ]
podman tag !$ quay.io/...
podman push !$
...and amended the calling tests accordingly.
I've tried to make the the smallest reasonable diff, not the
smallest possible one. I hope it's a reasonable tradeoff.
Signed-off-by: Ed Santiago <santiago@redhat.com>
podman-remote is in better shape now. Let's see what needs
to be done to reenable remote system tests.
- logs test: skip multilog, it doesn't work remote
- diff test: use -l only when local, not with remote
- many other tests: skip_if_remote, with 'FIXME: pending #xxxx'
where xxxx is a filed issue.
Unrelated: added new helper to skip_if_remote and _if_rootless,
where we check if the source message includes "remote"/"rootless"
and insert it if missing. This is a minor usability enhancement
to make it easier to understand at-a-glance why a skip triggers.
Signed-off-by: Ed Santiago <santiago@redhat.com>
Some CI tests are flaking in the SELinux test, possibly because
there's a new variation of the "multiple devices" warning I hadn't
seen before:
WARNING: Creating device "/dev/null" with same type, major and minor as existing "/dev/foodevdir/null".
Solution: in remove_same_dev_warning(), remove "multiple" from
the match string.
Also: fix a Go test that wasn't cleaning up after itself. And
add an actual test to it, not just check-exit-status.
Signed-off-by: Ed Santiago <santiago@redhat.com>
- Issue #6735 : problem with multiple namespaces; confirms
combinations of --userns=keep-id, --privileged, --user=XX
- Issue #6829 : --userns=keep-id will add a /etc/passwd entry
- Issue #6593 : podman exec, with --userns=keep-id, errors
(test is currently skipped because issue remains live)
...and, addendum: add new helper function, remove_same_dev_warning.
Some CI systems issue a warning on podman run --privileged:
WARNING: The same type, major and minor should not be used for multiple devices.
We already had special-case code to ignore than in the SELinux
test, but now we're seeing it in the new run tests I added, so
I've refactored the "ignore this warning" code and written
tests for the removal code.
Signed-off-by: Ed Santiago <santiago@redhat.com>
Reversion of one part of #6679: my handling of 'realpath'
would not work when $PODMAN is 'podman-remote --url etc'.
Trying to handle that case got unmaintainable; so instead
let's just force 'make {local,remote}system' to invoke
with a full PODMAN path. This breaks down if someone
runs the tests with a manual 'bats' invocation, but I
think I'm the only one who ever does that.
Since podman path will now be very long in the logs,
add code to logformatter to abbreviate it like we do
for the ginkgo logs.
And, one thing that has bugged me for a long time:
in the error logs, show a different prompt ('#' vs '$')
to distinguish root vs rootless. This should make it
much easier to see at-a-glance whether a log file
is root or not. Add tests for it.
Signed-off-by: Ed Santiago <santiago@redhat.com>
- rm: confirm 'rm' and 'rm -f' on running container
- build: shotgun test of workdir, cmd, env, labels
The new build test cd's to a temporary directory, which broke
test invocations using a relative path (./bin/podman). Added
code to detect relative paths and convert them to absolute.
Signed-off-by: Ed Santiago <santiago@redhat.com>
The systemd unit test never ran in CI and was broken for various
reasons. Fix the test to execute Podman in systemd units and to also
run generated units files.
Note: more tests will be added in the future. The simple check for now
will prevent regressions.
Signed-off-by: Valentin Rothberg <rothberg@redhat.com>
the current implementation of info, while typed, is very loosely done so. we need stronger types for our apiv2 implmentation and bindings.
Signed-off-by: Brent Baude <bbaude@redhat.com>
- run: --name (includes 'podman container exists' tests)
- run: --pull (always, never, missing)
- build: new test for ADD URL (#4420)
- exec: new test for issue #4785 (pipe getting lost)
- diff: new test
- selinux (mostly copied from docker-autotest)
Plus a bug fix: the wait_for_output() helper would continue
checking, eventually timing out, even if the container had
already exited (probably because of an error). Fix: as
part of the loop, run 'podman inspect' and bail out if
container is not running. Include exit code and logs.
Signed-off-by: Ed Santiago <santiago@redhat.com>
As initially written the test does not work other than in
a CI environment because it relies on an empty tag history.
Rewrite so we can guarantee that, by creating a new image.
Also add slightly more helpful tests: the initial tests
would just show "expected 0, got 1" which is unhelpful.
Tweak so we test on actual history contents, which will
show more informative messages on failure.
And, finally, clean up after ourselves.
Signed-off-by: Ed Santiago <santiago@redhat.com>
podman exec leaks an exec_pid_<hash> file for every exec in tmpfs,
it's known rhbz#1731117, this case makes sure leakage issue has
been fixed.
rhbz: https://bugzilla.redhat.com/show_bug.cgi?id=1731117
Signed-off-by: Alex Jia <chuanchang.jia@gmail.com>
podman-remote rm now works; that's the only thing we were
waiting for to enable podman-remote (varlink) system tests.
Add a (too-complicated, sorry) Makefile target that will
define a random socket path, start the podman varlink server,
and run the test suite using podman-remote.
Also: add two convenience functions, is_rootless and is_remote,
and use those in skip_if_rootless/if_remote and elsewhere
Also: workarounds for broken tests:
- basic version test: podman-remote emits an empty 'Client'
line. Just ignore it.
- looks like 'podman-remote pod' doesn't work; skip test.
Also: minor documentation update
Signed-off-by: Ed Santiago <santiago@redhat.com>
Fedora CI tests are failing on rawhide under kernel
5.3.0-0.rc1.git3.1.fc31 (rhbz#1736758). But there's
another insidious failure, a 4-hour hang in the
rootless tests on the same CI system. The culprit
line is in the podman build test, but it's actually
BATS itself that hangs, not the build command -- which
suggests that it's the usual FD 3 problem (see BATS README).
It would seem that podman is forking a process that
inherits fd 3 but that process is not getting cleaned
up when podman crashes upon encountering the kernel bug.
Today it's podman build, tomorrow it might be something
else. Let's just run all podman invocations in run_podman
with a non-bats FD 3.
Signed-off-by: Ed Santiago <santiago@redhat.com>
By default, podman points PIDFile in generated unit file to non-existent
location. As a result, the unit file, generated by podman, is broken:
an attempt to start this unit without prior modification results in a crash,
because systemd can not find the pidfile of service's main process.
Fix the value of "PIDFile" and add a system test for this case.
Signed-off-by: Danila Kiver <danila.kiver@mail.ru>
podman-remote now supports rm! That's what we needed to start
running BATS tests.
Although most tests don't actually work, some do, and maybe
the rest will start working over time. For now, disable them.
The only significant difference found is that podman-remote
strips fractional seconds from timestamps in JSON output.
Probably not something worth caring about.
Signed-off-by: Ed Santiago <santiago@redhat.com>
- document a recommended convention for fail-fast tests
- document the requirement for jq. (And, add a fail-fast
test for its presence; remove the duplicated checks
in subtests)
- add further sanity checks to 'help' test. Add missing
documentation. Remove a no-longer-needed workaround for
usage-message bug fixed in #2486
- add a documented TEMPLATE
- and, since we're at 1.1, enable 'Remote API' check in
version test
- better diagnostics in setup/teardown; add vim filetype hint;
better formatting of actual-vs-expect errors
- new pod-top, logs, build tests
- improve error messages
- add $IMAGE alias for ridiculous $PODMAN_TEST_IMAGE_FQN
- final cleanup, in prep for merge
Signed-off-by: Ed Santiago <santiago@redhat.com>
Initial attempt at writing a framework for podman system tests.
The idea is to define a useful set of primitives that will
make it easy to write actual tests and to interpret results
of failing ones.
This is a proof-of-concept right now; only a small number of
tests, by no means comprehensive. I am requesting review in
order to find showstopper problems: reasons why this approach
cannot work. Should there be none, we can work toward running
these as gating tests for Fedora and RHEL8.
Signed-off-by: Ed Santiago <santiago@redhat.com>
Ed Santiago
Sascha Grunert
Valentin Rothberg
Brent Baude
Alex Jia
Danila Kiver