The previous commit introduced a KeyError crash in tests_for_source() for
packages which are unbuilt/uninstallable and only present in unstable.
Ignore these in tests_for_source() as they can't possibly be a regression for
their dependencies, and there is no sensible way to run a test for them.
Commit 463 ("Don't promote packages with unbuilt reverse dependencies") turned
out to be too strict: This holds up too many innocent packages in -proposed.
If unstable has an unbuilt/uninstallable reverse dependency D of a package P,
trigger a test anyway (which will then most likely run against the testing
version of D). If that succeeds, the unstable P did not break D and can be
accepted. If it fails, D needs to be fixed.
Ideally we would set up some clever apt pinning to force installation of
testing-D, to avoid running into the uninstallability of unstable-D, but this
is tricky and error prone.
Drop the temporary "UNINST" state from commit 466 again. Instead, excuses.html
will now show a test against the testing version of D together with a note that
the unstable version is unbuilt/uninstallable.
This should ideally clear up all cases where a requested result is neither
present or pending. Log an error if that still happens (will be checked in the
next couple of runs), and ensure in the tests that we don't trigger any
outstanding "FIXME" log messages.
Commit 463 introduced waiting on reverse dependencies which are not built or⎵
installable yet, but set their status as "RUNNING". This is confusing as there
is no actual test in progress yet.
Instead, set their status to a new UNINST value, displaying as⎵
"Unbuilt/uninstallable"
If a reverse dependency D of a package P is not built yet, then D will be in
"exclusions" as we can't sensibly run D's tests at that time. In that case,
don't just ignore the missing test result but consider D's test as "in
progress".
Note that this might lead to stalling an innocent P if a broken (FTBFS) D gets
uploaded at the same time. This can/should be handled by overrides if fixing
D isn't appropriate, but this is better than allowing P to break D in that
situation.
Swift results were considered for older versions of triggers instead of waiting
for results for the actual package/version that triggered a new test.
This broke due to two reasons:
* When evaluating the test results we need to check whether we have a result
for the tested package/version that got triggered by the current excuse, not
just for any older excuse.
* AutoPackageTest.fetch_swift_results() re-downloaded all results for a
package due to a wrong "marker" value: The marker needs to be the
complete object path, not just the timestamp suffix. This caused old test
results to be considered as "newer than the given marker".
Now that we look at autopkgtest results from swift we can drop the
adt-britney/lp:auto-package-testing code from autopkgtest.py.
Note that we still need it for boottest.py.
Adjust TestBoottestEnd2End.test_with_adt() for cloud results.
Change AutoPackageTest.results() to evaluate the Swift results instead of the
adt-britney ones.
TODO:
- Add more tests (like for adt-britney)
- Drop triggering of adt-britney tests
- Drop adt-britney tests (which fail now)
- Adjust TestBoottestEnd2End.test_with_adt
Now that we look at autopkgtest results from swift we can drop the
adt-britney/lp:auto-package-testing code from autopkgtest.py.
Note that we still need it for boottest.py.
Change AutoPackageTest.results() to evaluate the Swift results instead of the
adt-britney ones.
TODO:
- Add more tests (like for adt-britney)
- Drop triggering of adt-britney tests
- Drop adt-britney tests (which fail now)
Add bool whether there is any successful test of src/arch of any version. This
will be used for detecting "regression" vs. "always failed".
WARNING: This changes the results.cache format, so results.cache has to be
removed and recreated before deploying this.
Commit 446 only considered a package's own tests. But we also need to check for
newer results of failed reverse dependency tests. Introduce a new
failed_tests_for_trigger() helper which computes the failed (src, arch) failed
tests for a given package, and fetch new results for all of them.
When collecting results, not only check pending tests, but also new results for
failed tests. This picks up new test results from manual retries which might
now have succeeded.
Their default values are invalid and must be set locally. But as
britney1-ubuntu copies these into production, we would run with an invalid
config with an unmodified config file.
Until now, autopkgtest results were triggered via an external "adt-britney"
command from lp:auto-package-testing. This required a lot of state files and
duplicated effort, uses hardcoded absolute paths to these external tools, and
is quite hard to understand and maintain. We also want to move away from
Jenkins and rsyncing state files.
Directly retrieve autopkgtest results from a publicly readable and browsable
Swift container, with a debci-compatible layout
(https://wiki.debian.org/debci/DistributedSpec). This now tracks both requests
and results on a per-architecture granularity, so that we can track
per-architecture regressions/always-failed.
Introduce a new ADT_SWIFT_URL config option that sets the swift base URL. If
this key is not set, the behaviour does not change compared to previous
versions, and no results will be retrieved from the cloud.
This still keeps the old adt-britney requests/results as the authoritative
data and for now merely shows the swift results in addition. With that we can
compare the results and run the cloud testing in parallel to find/fix problems
until we switch over. Due to that, the code to britney.py is temporary, does
*not* use AutoPackageTest.results(), and instead just reads the internal
results map.
This is necessary so that we can properly match requested to received results
when the latter arrive in different runs for different architectures.
This also opens up the possibility of per-arch blacklisting later.
Until now, autopkgtests were triggered via an external "adt-britney" command
from lp:auto-package-testing. This duplicated a lot of effort (such as its own
chdist and reverse dependency calculation), maintains a lot of state files,
uses hardcoded absolute paths to these external tools, and is quite hard to
understand and maintain. We want to get rid of all this.
Add logic to AutoPackageTest.request() to use britney's existing reverse
dependency maps and figure out the set of tests to run for packages in
unstable. These are being tracked in "requested_tests".
Add logic to AutoPackageTest.submit() to send test requests to the AMQP server
specified in the new "ADT_AMQP" config key. For testing this can be a file://
URL, and if not set no test requests will be sent at all.
The set of tests which were requested in previous runs are tracked in
UNSTABLE/autopkgtest/pending.txt, so that we don't re-request tests in
subsequent runs.
There is no implementation for collect() and results() yet, these will be done
in a separate commit.
Add initial test cases.
Don't change britney.conf from the source tree and restore it back after the
tests; this is prone to leave a broken config behind if a test gets
interrupted. Instead, copy the file to our temp dir, and let tests hack on it
there.
(Introduced in lp:~canonical-ci-engineering/britney/boottesting-support)
Extend read_sources to store a new AUTOPKGTEST boolean flag, which is true if
the Testsuite: field exists and starts with "autopkgtest" (this covers autodep8
cases like autopkgtest-pkg-perl).
Extend TestData.add() to take a new testsuite argument which specifies the
source's Testsuite: field.
