The issue with an implicit rollback is that the caller has no idea
whether it should invoke rollback or not if an exception occurs.
Signed-off-by: Niels Thykier <niels@thykier.net>
This isolates the undo handling in the new transaction object and in
doop_source, which currently generates the undo items. This commit
will be a stepping stone to rewriting the undo handling.
Signed-off-by: Niels Thykier <niels@thykier.net>
Support for binNMUs from *pu is currently broken, so disable it for now.
see https://bugs.debian.org/916209 for more info.
Signed-off-by: Ivo De Decker <ivodd@debian.org>
When building the excuses, don't calculate smoothbins for items not from
unstable.
This improves the change from commit db584d9fdc
Also try to make the code more readable.
Signed-off-by: Ivo De Decker <ivodd@debian.org>
When building the excuses, don't calculate smoothbins if the source is not in
unstable. Doing so would result in a KeyError and a crash. This happens for
sources that are in testing and (testing-)proposed-updates, but not in
unstable.
Signed-off-by: Ivo De Decker <ivodd@debian.org>
Support blocking the removal of a package with a removal hint. This is useful
to block an auto-removal.
Before this patch, only removals for packages not in unstable could be
blocked.
Signed-off-by: Ivo De Decker <ivodd@debian.org>
This commit updates the test suite to use the BinaryPackageUniverse
instead of the InstallabilityTester where that makes sense. The rest
of Britney has yet to be updated except where absolutely necessary (as
that will come in a later commit).
Signed-off-by: Niels Thykier <niels@thykier.net>
Don't make dependencies of cruft a blocker for being a candidate. In most
cases, the cruft won't migrate to testing, so it shouldn't be a blocker. If
the cruft would be considered for migration to testing, the installability
check will catch missing deps.
Signed-off-by: Ivo De Decker <ivodd@debian.org>
Signed-off-by: Niels Thykier <niels@thykier.net>
At the moment, britney will always read the nuncache from disk
(possibly regenerating it before reading it). However, we can save
the "read from disk" by moving the initialization of nuninst into the
constructor of Britney and relying on clone_nuninst.
Signed-off-by: Niels Thykier <niels@thykier.net>
Previously, we split affected into a "reverse relations with a
dependency on the migrated/change package" set and a set of "the
rest". With this commit we change the sets into a set of packages
with a direct relation on the migrated/changed package and the full
set of all affected.
The change has no effect on Britney's results as the first set is only
used as an optimization to reject some items faster.
Signed-off-by: Niels Thykier <niels@thykier.net>
At the moment, the two are identical, but they are not intended to
remain identical (i.e. the name should eventually become the full
suite name).
Signed-off-by: Niels Thykier <niels@thykier.net>
Cherry-Pick: 80bf9060de
Cherry-Pick: f32907acea
Cherry-Pick: 9ef496177f86b18d9f910da1360dd773b82f1fb7
Cherry-Pick: b16530a37d
Signed-off-by: Niels Thykier <niels@thykier.net>
Flatten the defaultdict(set) for unsat_deps into a standard dict for output
When determining whether a policy applies to a given item, use the
suite class rather than the suite name.
Signed-off-by: Niels Thykier <niels@thykier.net>
The "almost" comes from the fact that there is a reference to
"pu"/"stable" which does not have an obvious fix at the moment.
Signed-off-by: Niels Thykier <niels@thykier.net>
Previously, we would always create empty data structures for "pu" and
"tpu" if these optional suites were not configured. However, we only
reference "pu" and "tpu" directly in very few spots that can easily be
rewritten.
Signed-off-by: Niels Thykier <niels@thykier.net>
Into 3 categories:
* target suite ("testing")
* primary source suite ("unstable")
* additional source suites ("pu" and "tpu")
This will be useful for implementing logic working with suites without
basing it on the name of the suite.
Signed-off-by: Niels Thykier <niels@thykier.net>
At the moment, it is just a glorified dict. However, we will
eventually use it to get rid of the hardcoded references to "testing"
etc. all over the code.
Signed-off-by: Niels Thykier <niels@thykier.net>
All the output is a duplicate of what is being sent to the root logger
(albeit in a different format) and in e.g. our test suite the output
is entirely redundant.
Signed-off-by: Niels Thykier <niels@thykier.net>
With a bit of code we can replace the manual file-handling for
"upgrade_output" with a logger. This will enable us to refactor other
bits that currently depend on "output_write" by making those bits use
a logger instead.
This also migrates "do_hint" to use the new output logger. This is
due to "do_hint" being the only method relying on writing of partial
lines and maintaining support for that in "output_write" would have
been non-trivial.
To ensure "pretty" formatting to stdout, the messages in
"output_write" are now chopped into multiple lines.
The only visible change is that the output to stdout from
"output_write" now also includes the prefix with a timestamp.
However, then contents of "upgrade_output" remain unchanged
deliberately.
Signed-off-by: Niels Thykier <niels@thykier.net>
We already know the item is successful when we print it and the next
line ("final:") will confirm it any way.
Signed-off-by: Niels Thykier <niels@thykier.net>
At a glance, it looks like the value of "better" variable can be
decided from 3-4 places. However, due to the code flow only two of
those assignments are truly "live"/useful.
Signed-off-by: Niels Thykier <niels@thykier.net>
This commit rewrites the make-shift "log" methods to use the logging
framework without requiring changes to the callers. This will be done
in a latter commit to keep things reviewable.
Signed-off-by: Niels Thykier <niels@thykier.net>
The "Out of date" binaries loop has gotten too complex to also handle
the "unsatisifiable dependency" check. Concretely, we failed to
generate proper excuses for arch:all packages due to this.
Separate the two loops to restore the arch:all check.
Signed-off-by: Niels Thykier <niels@thykier.net>
Add a new "BuildDependsPolicy" that will check the satisfiability of
the build-dependencies listed in the Build-Depends and
Build-Depends-Arch fields. This enables gating of packages based on
missing / broken build-dependencies.
There are some limitations:
* Build-Depends-Indep is ignored for now. Missing or broken packages
listed in Build-Depends-Indep will be continue to be silently
ignored.
* Being a policy check, it does not enforce "self-containedness" as
a package can still migrate before a build-dependency. However,
this can only happen if the build-dependency is ready to migrate
itself. If the build-dependency is not ready (e.g. new RC bugs),
then packages build-depending on it cannot migrate either (unless
the version in testing satisfies there requirements).
Signed-off-by: Niels Thykier <niels@thykier.net>
We basically use them as sets and do not need to rely on the ordering,
so we might as well just turn them into proper sets.
Signed-off-by: Niels Thykier <niels@thykier.net>
Niels Thykier
Ivo De Decker
Mathieu Trudel-Lapierre
Paul Gevers