The autopkgtest policy prints when it doesn't run a test due to a
package being unsatisfiable on an arch. This can be a bit mysterious,
and it can seem like it's causing the package to not become a candidate,
when it's not. Print some info when we decide not to skip indicating
what happened.
We currently concatenate all triggers together into a string, but the
AMQP consumer expects this to be a list.
When using AMQP, keep the triggers as a list. Ensure that the "real"
trigger (the package being tested) is kept first, as before.
On some distros (Ubuntu), arch:all packages are built along with one of
the architectures. We shouldn't be listing 'all' as its own arch in this
case. Instead we filter out the binaries except for on the
'all_buildarch'.
We just had the autopkgtest queues DoSed because britney was crashing
after requesting each reverse dependency for a perl upload, but before
it had written pending.json out so it knew what not to request again.
This was 25,000 requests per arch...
Let's write pending.json straight after sending each request, so that
the next run - even after a crash - won't re-request the same things
again.
These have a hash appended.
We don't actually use the baseline retrying, which is where the ID
parsing is used, but we might as well handle this, not least so we don't
crash.
The apt version comparison sorts 'blacklisted' greater than most version
numbers, which means that we accidentally apply force hints for version
'blacklisted' to all uploads. Since this is the only case of a hacked
version number, let's special case it so that 'blacklisted' hints only
match packages with 'blacklisted' version.
We're supposed to synthesise an "unknown" version for these, but a bug
in the worker meant we didn't do that in some cases and these leaked
into swift. Let's repair it client-side.
Most DKMS packages do not declare Testsuite: autopkgtest-pkg-dkms, but
we can detect this anyway, and this way we can enforce that the module
is buildable.
It works like this. We wait until all tests have finished running. and
then grab their results. If there are any regressions, we mail each bug
with a link to pending-sru.html. There's a state file which records the
mails we've sent out, so that we don't mail the same bug multiple times.
We want to treat linux-$flavor and linux-meta-$flavor as one set in
britney which goes in together or not at all. We never want to promote
linux-$flavor without the accompanying linux-meta-$flavor.
Introduce a synthetic linux* → linux-meta* build-dependency to enforce
this grouping.
For packages with lots of reverse dependencies, new versions of those reverse
dependencies may keep on showing up in testing. If migration is blocked until
the results for these new version, migration may take extremely long. If there
are results for the current trigger but for the previous version of the reverse
dependency, use those until the fresh resuts are available.
Similar for the reference runs.
Hint to allow smooth update, even if the section isn't allowed in the
configuration.
Note that this takes the source name and the source version IN TESTING
of the binaries that must be allowed to stay around to allow a smooth
update.
Currently, britney only schedules reference runs when they don't exist. It does
strip out runs against older versions of the autopkgtest, but the current version
may exist for a while and the reference run can be old. So, add an option to
ignore old results.
We only want to add packages which conflict in testing, but don't
conflict in unstable. For those, the newer version (from unstable)
probably fixes the conflict.
When an arch:all binary is uninstallable on a non-nobreakall arch, don't
block it, but still mark it as uninstallable, so the autopkgtest policy
knows not to schedule tests for it.
Currently when a package is uninstallable on an arch, no autopkgtests for that arch are triggered
and the autopkgtest policy blocks migration. However it's not the job of the autopkgtest policy
to judge uninstallability and packages that build an arch:all package that just isn't installable
on the autopkgtest arch should not be blocked for this.
Closes: #918620
Rework the dependencies between excuses.
Dependencies are specified based on packages (source or binary). Based
on these packages, dependencies on other excuses (that have these
packages) are calculated.
As with package dependencies, dependencies between excuses can contain
alternatives.
Each alternative can satisfy the dependency, so the excuse only becomes
invalid if all of the alternatives of a specific dependency are
invalidated.
Tracking of the alternatives and their validity is moved to separate
objects.
Because autopkgtest failure in a package is now a serious bug (RC), packages
that don't opt-in should not be tested. Due to the way autopkgtest and autodep8
work together, package that don't declare they have an autopkgtest can still be
tested. However, maintainers should not be force to say their package doesn't
work with autodep8 by introducing a fake test.
When there is no test in testing, and a failing test in unstable, don't
allow the package to migrate. Doing so would make it instantly RC-buggy.
Signed-off-by: Ivo De Decker <ivodd@debian.org>
Iain Lane
Iain Lane
Łukasz 'sil2100' Zemczak
Tiago Daitx
Robert Bruce Park
Colin Watson
Martin Pitt
Paul Gevers
Ivo De Decker