The passthrough_env function was unnecessarily complicated,
hence fragile. Clean it up, and add regression tests.
For future reference: CI broke horribly because of this.
Rootless tests all failed with missing CI_DESIRED_NETWORK.
Root cause was that CIRRUS_CHANGE_TITLE had a trailing
space which, because of shell indirection, passthrough_env()
wrote as trailing backslash (not backslash-space) in the
/etc/ci_environment file, which then caused the next line
in the file to get glommed onto CIRRUS_CHANGE_TITLE.
Signed-off-by: Ed Santiago <santiago@redhat.com>
...not as a separate post-main step. Reasons:
1) If main test times out, Cirrus won't run subsequent steps
2) It really belongs in the main log anyway, because when
looking at test results, you want to have instant access
to versions, you don't want to have to clickety-click
ten steps.
Signed-off-by: Ed Santiago <santiago@redhat.com>
Intended to be run from nightly Cirrus cron job.
1) Queries github for highest-sorting (not necessarily "latest") tag
2) Checks that the Windows MSI exists, fails if not
3) Cross-checks markdown files to ensure they have up-to-date links
When run interactively, it will auto-update the .md files
to show and link to the latest version. This makes it easy
for anyone to then submit an update PR.
And, it turns out that MSI is obsolete, the new thing is EXE.
Update the tutorials to reflect that.
Signed-off-by: Ed Santiago <santiago@redhat.com>
All the other Windows tasks depend on access to a podman-remote build
from the Alt. Arch. `Windows Cross` task. Re-arrange the test-skipping
call to never skip here only.
Signed-off-by: Chris Evich <cevich@redhat.com>
We've had some oopsies in system tests:
podman foo bar
run podman foo bar
...all of which should be run_podman with underscore. Those
have been passing because /usr/bin/podman is the fallback
from $PATH. In those (few) cases, we haven't actually been
testing the podman we should be testing.
Solution: nuke /usr/bin/podman and podman-remote before
invoking system and unit tests. As an extra level of
paranoia, check for other podmans in $PATH - if any
exist, bail out with a fatal error.
Also: in a few cases where runner.sh invokes podman for
containerized something-something, run bin/podman instead
of podman from $PATH.
Also: fix existing dependencies on /usr/bin/podman
Signed-off-by: Ed Santiago <santiago@redhat.com>
It's conceivable for CI to spend a lot of time testing code which
otherwise should be rejected due to quality problems. Previously this
was validated in a dedicated task, however a failure would still fail
the CI run. Simplify the number of CI tasks by combining the consistency
check at the tail-end of the build task.
Signed-off-by: Chris Evich <cevich@redhat.com>
Previously, two tasks always ran first, prior to anything else. One to
verify network and external-service connectivity. Another to verify
certain important `.cirrus.yml` standards are met. However, as the
total number of tasks continues to grow, the need to keep these basic
checks as dedicated prerequisites is of decreasing value/importance.
Fold these two checks into a new `pretesting_script` component of the
Fedora `build` task, on both `x86_64` and `aarch64`.
Signed-off-by: Chris Evich <cevich@redhat.com>
At the time of this commit, there's no easier way to detect this using
`golangci-lint` or the go tool (that I could find). A future update
to the `go list` command may support detection, for now use a CI script.
Signed-off-by: Chris Evich <cevich@redhat.com>
new file: test/e2e/config_arm64.go
Tests that fail on aarch64 have been skipped with
`skip_if_aarch64`.
Co-authored-by: Chris Evich <cevich@redhat.com>
Co-authored-by: Ed Santiago <santiago@redhat.com>
Signed-off-by: Lokesh Mandvekar <lsm5@fedoraproject.org>
...and enable the at-test-time confirmation, the one that
double-checks that if CI requests runc we actually use runc.
This exposed a nasty surprise in our setup: there are steps to
define $OCI_RUNTIME, but that's actually a total fakeout!
OCI_RUNTIME is used only in e2e tests, it has no effect
whatsoever on actual podman itself as invoked via command
line such as in system tests. Solution: use containers.conf
Given how fragile all this runtime stuff is, I've also added
new tests (e2e and system) that will check $CI_DESIRED_RUNTIME.
Image source: https://github.com/containers/automation_images/pull/146
Since we haven't actually been testing with runc, we need
to fix a few tests:
- handle an error-message change (make it work in both crun and runc)
- skip one system test, "survive service stop", that doesn't
work with runc and I don't think we care.
...and skip a bunch, filing issues for each:
- #15013 pod create --share-parent
- #15014 timeout in dd
- #15015 checkpoint tests time out under $CONTAINER
- #15017 networking timeout with registry
- #15018 restore --pod gripes about missing --pod
- #15025 run --uidmap broken
- #15027 pod inspect cgrouppath broken
- ...and a bunch more ("podman pause") that probably don't
even merit filing an issue.
Also, use /dev/urandom in one test (was: /dev/random) because
the test is timing out and /dev/urandom does not block. (But
the test is still timing out anyway, even with this change)
Also, as part of the VM switch we are now using go 1.18 (up
from 1.17) and this broke the gitlab tests. Thanks to @Luap99
for a quick fix.
Also, slight tweak to #15021: include the timeout value, and
reword message so command string is at end.
Also, fixed a misspelling in a test name.
Fixes: #14833
Signed-off-by: Ed Santiago <santiago@redhat.com>
It was not obvious enough in the scripts how much of a snowflake this
environment is. Fix that with lots of capitalized words and asterisks.
Signed-off-by: Chris Evich <cevich@redhat.com>
We're still not testing runc in CI (#14833), and it may be weeks
or months before we can, due to criu/glibc nightmare, but one day
we'll be back on track, then later on we'll update VMs again,
and screw it up, and lose runc, and not notice, and RHEL will
break, and oh noes headless chicken again, repeat repeat.
We can do better. Use .cirrus.yml to explicitly define which
VMs should use which runtimes, and enforce it early in the
CI build step. This should never fail (uh huh) in a PR,
only in one of the update-VM PRs.
Signed-off-by: Ed Santiago <santiago@redhat.com>
The podman-machine integration tests are designed to execute on
bare-metal, since they perform significant work with virtual-machines.
This test is costly to run at scale, so it is limited to being manually
triggered by developers (for now). A 'trigger' button will appear in the
task status page of the Github WebUI once all test dependencies are met.
In the Cirrus-CI WebUI, there is also a 'pre-trigger' button that may be
pressed if a developer doesn't wish to wait. Also:
* Add a `localmachine` target in the `Makefile` on the off-chance
developers wish to execute locally. Update the `ginkgo-run` target
to accommodate re-use by the new `localmachine` target.
* Exclude `podman_machine` task from `success` dependency verification.
This also involves adding an exception to `cirrus_yaml_test.py`
otherwise it will complain loudly.
* ***NOTE*** Inclusion of `ec2_instance` in *any* task will cause
`hack/get_ci_vm.sh` to barf and be non-functional. Future updates will
be made to restore functionality. Before then, simply comment out
the `ec2_instance` section as a temporarily workaround.
Signed-off-by: Chris Evich <cevich@redhat.com>
One day we may use AWS for part of CI. Do you want to maintain
two separate code paths in this script for linking to artifacts
in multiple cloud providers? Can you say no? I knew you could.
Cirrus already knows the location of the artifacts and provides
a transparent mechanism for accessing them. Use it.
This PR exposed a nasty bug in our environment-variable handling:
envariables passed through to the containerized environment were
being double-space-escaped, so "FOO=a b" ended up as "FOO=a\ b"
(with a backslash), with one consequence being invalid URLs.
The solution is simple: run 'podman -e FOO', not '-e FOO=value'.
Finally, reinstate the environment-variable dump (in comments).
I had removed this in a moment of panic over leaking secrets,
but no, that doesn't happen. Exclude scary-sounding vars anyway.
Signed-off-by: Ed Santiago <santiago@redhat.com>
Fixes#14021
Substitution values built from `$(shell ...)` output can easily be empty
due to the shell's default `pipefail` behavior. This can also hide
non-zero exit codes, similarly resulting in empty values being set.
While not a perfect fix, the situation is improved by using the
`err_if_empty` function in all cases where empty values would be
unexpected. Remove the definitions for `GIT_BRANCH` and
`GIT_BRANCH_CLEAN` which don't seem to actually be used anywhere
(including in code).
Add a simple release-test to verify `podman info` outputs a non-empty
value for "GitCommit".
Signed-off-by: Chris Evich <cevich@redhat.com>
Most were simply deleted, the main one addressed is in the
"pre-testing" `ext_svc_check.sh` script. It will now verify
accessibility of several key test images we maintain in `quay.io`.
Signed-off-by: Chris Evich <cevich@redhat.com>
Followup to #14178: I broke CI. "--fork-point" does not do
what I thought it did, so remove it.
Also, add 'set -x' to help us debug the next time something
like this happens.
Signed-off-by: Ed Santiago <santiago@redhat.com>
The use of 'C^' (parent) in 'git rebase' is counterintuitive,
at least to me: when C is a merge of multiple commits, rebase
picks each of those commits. That probably makes good sense
to a git expert, which I'm not.
Solution: forget using '^', just calculate the baseline sizes
by doing a checkout of the PR's parent.
Also: compute PR parent using git-merge-base instead of
blindly trusting $DEST_BRANCH (which may be volatile as
other PRs are merged).
Also: run git-rebase directly, not via make rule. That
indirection is too confusing here.
Signed-off-by: Ed Santiago <santiago@redhat.com>
The test-skipping optimization is failing as rootless on non-main,
because $DEST_BRANCH is not set. Solution: check for envariable,
skip test if missing. (This was part of my original PR, but was
accidentally removed in #14013)
Also: DEST_BRANCH was silently being defaulted to 'main' in lib.sh.
Remove that: per @cevich, it is no longer necessary.
Fixes: #14131
Signed-off-by: Ed Santiago <santiago@redhat.com>
Newer versions of git are much more pedantic about who owns the
repository files. When setting up to run rootless, prior to this
commit, the repo. ownership was changed from root. This causes
all subsequent git-operations as root to fail:
```
fatal: unsafe repository ('<$GOSRC>' is owned by someone else)
```
Fix this by re-ordering operations, such that the change in ownership is
done immediately before executing as a user. Also disable the
git-ownership check on the source repository assuming the CI environment
is disposable.
Signed-off-by: Chris Evich <cevich@redhat.com>
Now that netavark and aardvark are packaged and default in F36, support
CNI-based testing in F35 and Ubuntu.
* Remove the temporary/special `$TEST_ENVIRON=host-netavark` construct.
* Remove dedicated/special integration and system testing tasks.
* Update test-config setup to properly handle CNI vs netavark/aardvark
environments.
* Update package-version logging to operate based on installed packages
(along with some other minor script cleanups).
* Update global environment setup to force `$NETWORK_BACKEND=netavark`
in F36 and later. Except when `upgrade_test` task runs.
* Discontinue installing netavark and aardvark-dns binaries from
upstream build artifacts.
* Drop CGV1-vs-2 policy check. Ubuntu VMs now exclusively test CGv1,
Fedora VMs test CGv2, with F35 testing CNI and F36 testing Netavark.
Signed-off-by: Chris Evich <cevich@redhat.com>
The originally intent for skipping tests based on change-content was to
optimize the PR workflow. However, a mistake in a conditional is
causing almost all tasks running for Cron and branches to be skipped.
Fix this by checking for an empty '$CIRRUS_PR' variable. This value is
always empty when operating outside of PRs.
Signed-off-by: Chris Evich <cevich@redhat.com>
A suspected race on uploading gopath cache is causing the docker-py (and
possibly other) tasks to fail unpredictably with an error from `make`
regarding missing `bin/podman`. Since this failure is affecting all
development activity, apply a quick/dirty fix to the failing task, by
simply rebuilding the binary. A more comprehensive/long-term fix will
be worked in a future PR.
Signed-off-by: Chris Evich <cevich@redhat.com>
The default verbosity level does not show the classes or function names.
This makes it difficult to debug problems like hangs. Also, separate
the bats and python-based tests into two sections. This allows for
easier debugging, since isolation can be done in `runner.sh` rather than
mucking with the `Makefile`. Lastly, update the logformatter script
to `autoflush stdout` (thanks @edsantiago).
Signed-off-by: Chris Evich <cevich@redhat.com>
A common pattern is to submit PRs that update only tests or docs.
When the only changes are to test/e2e, there is no point in running
test/system or test/upgrade or test/buildah-bud. Likewise, reciprocally,
and similarly for a bunch of other tests (alt, cross, apiv2, ...)
And when the only changes are under docs/ , there is no point in
running any of the above.
Exception: if $CIRRUS_<mumble> are undefined (e.g., cron), never skip
Signed-off-by: Ed Santiago <santiago@redhat.com>
Add a CI check to prevent unwanted bloat in binary images,
by building a baseline (pre-PR) binary then comparing file
sizes post-PR.
Part 1 (#13518) added a new script that runs multiple 'make's,
comparing image sizes against an original, and failing loudly
if growth is too big. An override mechanism is defined.
This is part 2 of 2: adding the CI rule. We couldn't do that
in part 1, because the rule would call a script that didn't
exist in the pre-PR commit.
Signed-off-by: Ed Santiago <santiago@redhat.com>
Add a pair of new Cirrus test suites using Compose v2 instead of
Compose v1 (as is currently packaged in Fedora). They work
identically, and run the same tests, as the Compose v1 tests, but
with the new v2 binary instead.
[NO NEW TESTS NEEDED] This adds an entire Cirrus suite...
Signed-off-by: Matthew Heon <mheon@redhat.com>
* Add configuration to add report header for python client used in tests
* Move report headers into the individual test runners vs runner.sh
Signed-off-by: Jhon Honce <jhonce@redhat.com>
Building from source would involve separate `make` and `make install`
steps.
This removes a lot of unnecessary `-nobuild` targets which were
otherwise needed for packaging.
This commit also removes spec files for unused copr jobs.
[NO NEW TESTS NEEDED]
Signed-off-by: Lokesh Mandvekar <lsm5@fedoraproject.org>
This PR adds the CI mechanisms to obtain the latest upstream netavark
binary, and set a magic env-var to indicate e2e tests should execute
podman with `--network-driver=netavark`. A future commit implement
this functionality within the e2e tests.
Due to the way the new environment is enabled, the standard task name
is too long for github to display without adding ellipsis. Force the
custom task name `Netavark Integration` to workaround this. At some
future point, when netavark is more mainstream/widely supported, this
custom task and upstream binary install can simply be removed - i.e.
netavark will simply be used by default in the normal e2e tasks.
Signed-off-by: Chris Evich <cevich@redhat.com>
An error was observed in another PR while downloading the swagger
binary. The error was relating to the upstream egress quota. Obviously
our downloading it every time for each CI run isn't helping. Fix this
by moving the download into the image-build process, and simply re-use
the already present binary here.
Ref: https://github.com/containers/automation_images/pull/103
Signed-off-by: Chris Evich <cevich@redhat.com>
Add a magic 'echo' to runner.sh, displaying $GIT_COMMIT in
a special syntax. The logformatter script, seeing this,
will hyperlink error messages to the failing source file.
Signed-off-by: Ed Santiago <santiago@redhat.com>
During initial testing of Fedora 35beta VM images in CI, the bindings
task was timing out. In order to allow time for collection of system
details (logs), execution needs to timeout earlier than the task.
Under normal conditions, the bindings test finishes in about 10-minutes.
Use the ginkgo timeout option to limit execution, so it times out after
30 minutes.
Also add the `-progress` option so the output more closely resembles how
ginkgo runs the integration tests.
Signed-off-by: Chris Evich <cevich@redhat.com>
Mount a directory from /var/tmp to /tmp to make sure that /tmp is not on
an overlay mount. This should make overlay mounts possible in the
containerized tests which we're currently skipping.
Signed-off-by: Valentin Rothberg <rothberg@redhat.com>
Add execution of the downstream gitlab-runner tests using
rootless podman through the magic of socket-level
docker compatibility. Include a comment suggesting how
to temporarily disable the test in case it fails beyond
podman code scope.
Signed-off-by: Chris Evich <cevich@redhat.com>
Fixes#11417
Cross-building the podman-remote documentation requires a functional
native architecture executable. However `make` only deals with
files/timestamps, it doesn't understand if an existing binary will
function on the system or not. This makes building cross-platform
releases incredibly accident-prone and fragile.
A practical way to deal with this, is via multiple conditional (nested)
`make` calls along with careful manipulation of `$GOOS` and `$GOARCH`.
Also, when cross-building releases be kind to humans and cleanup
any non-native binaries left behind.
Update the `Alt Arch. Cross` Cirrus-CI task to build release archives
for all Linux architectures supported by golang and podman. Update
the `OSX Cross` task to additionally build for the M1 (arm64)
architecture.
Finally, update the release process documentation to reflect the
new locations (Cirrus-CI task names) for the release archives. Include
a note about additional manual work being required to produce the
signed `.dmg` file for MacOS.
Signed-off-by: Chris Evich <cevich@redhat.com>
Significant bitrot results in almost immediate test failure. This
commit adds only the very basic, bare-minimum needed to get them
started.
***TESTING RESULTS ARE IGNORED***
Signed-off-by: Chris Evich <cevich@redhat.com>
The rootless cni namespace needs a valid /etc/resolv.conf file. On some
distros is a symlink to somewhere under /run. Because the kernel will
follow the symlink before mounting, it is not possible to mount a file
at exactly /etc/resolv.conf. We have to ensure that the link target will
be available in the rootless cni mount ns.
Fixes#10855
Also fixed a bug in the /var/lib/cni directory lookup logic. It used
`filepath.Base` instead of `filepath.Dir` and thus looping infinitely.
Fixes#10857
[NO TESTS NEEDED]
Signed-off-by: Paul Holzinger <pholzing@redhat.com>
The new cobra v1.2.0 release brings a number of bug fixes for shell
completion scripts. Regenerate the scripts with `make completions`
to sync them with the upstream version, currently we have some custom
ones to avoid some upstream bugs. Because the new cobra version has
all fixes we should use the upstream scripts.
Add a check to CI to ensure we always use the up to date scripts.
[NO TESTS NEEDED]
Signed-off-by: Paul Holzinger <pholzing@redhat.com>
* Incorporate changes from abandoned #9918: Use dedicated `bin`
sub-directories for `windows` and `darwin` when building
`podman-remote`. The linux flavor remains under `bin` as before.
* Fix MacOS Documentation-generation for release-packaging.
The `install-podman-remote-%-docs` target requires local execution
of `podman-remote`, but it was assuming GOOS=linux. Fix this
by dynamically discovering the local OS/architecture type while
still permitting cross-building of MacOS binaries under Linux.
* Unify temporary directory/file behavior to use a common template.
In case of left-over temporary items left in the repository,
update the `clean` target accordingly to remove them.
* Fix broken podman-remote-static and MacOS release archive targets
mismatching the `podman-remote-%` target. Disambiguate this target
for all platforms by spelling each out in full, instead of using
a wild-card recipe.
* Fix Windows-installer target to properly recognize existing
output files and not constantly rebuild every time.
* Include the podman version number in the Windows-installer target
in case a user downloads multiple releases.
* Include a subdirectory containing the podman version number for
both `tar.gz` and `zip` targets. This prevents users clobbering
existing directories when un-archiving from releases.
Signed-off-by: Chris Evich <cevich@redhat.com>
