This bug involves a corner case that involves:
* source orig providing liborig1 and orig-doc in testing
* source orig providing liborig2 and orig-doc in unstable
* source hijack providing liborig2 and orig-doc in both testing and unstable,
where the versions of hijack's binaries has a higher version than those of
"orig".
The arch:all packages are needed to trigger this, because Britney
flags an arch:any package as "out of date" and stops the migration
there. However, she is more lenient with arch:all packages.
What happens is that Britney realises that src:orig need to be updated
in testing (to remove liborig1). This leaves src:orig with no
binaries left in testing (as the orig-doc from hijack is used) and it
is therefore removed as an obsolete source. The obsolete removal then
exploded because Britney was also trying to remove the liborig1
package, which is no longer there.
Signed-off-by: Niels Thykier <niels@thykier.net>
With this patch, Britney will correctly parse (and deparse) a
versioned Provides. Furthermore, she will allow it to satisfy any
unversioned dependency on the provided package.
This is the easy half of #786803.
Signed-off-by: Niels Thykier <niels@thykier.net>
It can be used to A) to make the mismatch check more efficient and B)
share identical binaries between suites.
Signed-off-by: Niels Thykier <niels@thykier.net>
Previously whether such packages received excuses, and the specific
content of such excuses, was dependent on the order in which their
binary packages were considered.
Signed-off-by: Adam D. Barratt <adam@adam-barratt.org.uk>
This made iter_packages_hint a thin wrapper around try_migration, so
it was inlined into its only caller "do_all".
Signed-off-by: Niels Thykier <niels@thykier.net>
The callers of get_dependency_solvers need to do those table lookups
anyway. By moving it out, it is now possible to reuse the results.
Signed-off-by: Niels Thykier <niels@thykier.net>
Rely on the Installability tester to locate all of the affected
packages plus their transitive reverse dependencies. As the
InstallabilityTester is suite agnostic, the set of affected packages
now includes (versions of) packages not in testing, which is filtered
out during the check.
Signed-off-by: Niels Thykier <niels@thykier.net>
If we have an up-to-date arch all package available for this architecture,
that doesn't mean this architecture has an up-to-date build. We need an
architecture specific up-to-date package for that.
Signed-off-by: Ivo De Decker <ivodd@debian.org>
For fucked architectures, binaries from older versions are allowed to be in
testing, so we only remove them if they are gone from unstable.
Signed-off-by: Ivo De Decker <ivodd@debian.org>
Move nuninst cloning out of the check loop and always populate the new
nuninst entirely.
This will allow some simplifications in other places.
Signed-off-by: Niels Thykier <niels@thykier.net>
Use a set to filter out seen items to avoid doing O(n^2)
de-duplication. For very large hints, this can take considerable
time.
Using "seen_items" to build the actual hints on the (unverified)
assumption that Python can do something "smart" to turn a set into a
frozenset faster than it can with a list.
Signed-off-by: Niels Thykier <niels@thykier.net>
Britney is now smart enough to produce the same result from hints
regardless of the order of the items in the hint. With this in mind,
we can have the original auto-hinter produce hints as sets and filter
out duplicates as we produce them.
Note that the hints are sorted to produce deterministic output (to
make it easier to compare the hints between runs and changes).
Signed-off-by: Niels Thykier <niels@thykier.net>
Avoid some cases of O(n^2) behaviour in sort_actions and reduce the
size of n for the remaining O(n^2)-ish behaviour by filtering out
removals early on.
Signed-off-by: Niels Thykier <niels@thykier.net>
It's more natural to say "check this package if the current arch is in
this list" than "do not check this package if the current arch is not
in this list"
Signed-off-by: Adam D. Barratt <adam@adam-barratt.org.uk>
This updates the doop_source and _compute_groups functions so
that binary packages that are built from a different source
aren't included as part of an update to the original source.
In the event that it's a binary-only update, also don't remove
the hijacked packages from testing.
This change also removes an obsolete comment regardarding pre-conditions
for the _compute_groups function.
The live-data tests rely on an inconsistency, since they were before
Britney started to record the Essential field.
Signed-off-by: Niels Thykier <niels@thykier.net>
britney assumes that a package build is uniquely described by its
name, version and architecture. Particularly when constructing
Packages files by hand for testing purposes this assumption can be
violated, leading to confusing behaviour. This change makes britney
look for such mismatches, and report if any are found.
Notably:
* Avoid repeated calls frozenset(X), where we can trivially do
without.
* Skip the inner loop, when "i" is in "to_skip".
* Use a set rather than a list for "to_skip" as we do more
membership tests.
Signed-off-by: Niels Thykier <niels@thykier.net>
The get_dependency_solvers method returns a (boolean, list)-tuple, but
the boolean can always be implied from the list (in boolean context).
Signed-off-by: Niels Thykier <niels@thykier.net>
cmp is gone in python3. Also add a sorting method to Excuse that is
compatible with its __eq__/__hash__ methods.
Signed-off-by: Julien Cristau <jcristau@debian.org>
It doesn't exist in python3, but 1000 days should be safe enough as a
fallback for a package without urgency.
Signed-off-by: Julien Cristau <jcristau@debian.org>
- split the one-liner into a for and an if
- use open() as a context manager
- don't use string.strip which is gone in python3
Signed-off-by: Julien Cristau <jcristau@debian.org>
The "remark" hint is only intended for showing up in the output of "d"
(or via hint grep). It has no effect on Britney's behaviour.
Admittedly, the original code would have ignored it as well. But this
change makes it explicit and not simply a "ignored due to insufficient
permissions".
Signed-off-by: Niels Thykier <niels@thykier.net>
For out-of-date binaries, generate different excuses when the build is
missing, or when old (cruft) binaries for previous version are still around.
Signed-off-by: Ivo De Decker <ivodd@debian.org>
As part of a migration, we remove all the existing binaries built by
the source (possibly on a particular architecture) from testing; this
includes architecture-independent binary packages. However, when a
binNMU is in *pu, only the arch-dependent binary pakcages are present.
As a result, after the migration the architecture-independent packages
are no longer present in testing. This usually isn't a practical
problem, as dak will re-add them when it generates the packages files.
It is, however, wrong and will break if a source migration is tempted
during the same run as (and after) the *pu binary migration happened.
The simple fix is to not remove the architecture-independent packages
when performing such migrations.
Signed-off-by: Adam D. Barratt <adam@adam-barratt.org.uk>
In the rare case that a hint removed an uninstallable binary, the
binary could still be included in the nuninst counter.
Regression introduced in a46dd88.
Signed-off-by: Niels Thykier <niels@thykier.net>
sort_actions() can be quite expensive and it is wasteful to resort
actions after each successful "easy"-hint.
Signed-off-by: Niels Thykier <niels@thykier.net>
There are no uses of "lundo" left for a non-hint recurse run (i.e.
the "main run"), so there is no point in building it.
The "lundo"-list is still used in the recurse run of a "hint"-hint.
Signed-off-by: Niels Thykier <niels@thykier.net>
The "do_all"-method now checks the architectures of all changes
applied. If they entirely consist of items from "break archs", then
"do_all" will disregard the current "break archs" setting when
comparing nuninst counters.
This change avoids unintended installability regressions on break
arches when a hint (manual or automatic) apply only to packages on
break arches.
Signed-off-by: Niels Thykier <niels@thykier.net>
Rename local variables and avoid repeated chained lookups. In
particular, avoid confusing cases like:
[...]
version = binaries[parch][0][binary][VERSION]
[...]
binaries[parch][0][binary] = self.binaries[item.suite][parch][0][binary]
version = binaries[parch][0][binary][VERSION]
Where "version" here will refer to two different versions. The former
the version from testing of a hijacked binary and the latter the
version from the source suite (despite the look up using the "testing"
table, due to the testing copy being updated).
Notable renamings:
* binaries => packages_t (a.k.a. self.binaries['testing'])
* binaries[parch][0] => binaries_t_a
* binaries[parch][1] => provides_t_a
* Similar naming used for "item.suite" instead of "testing"
The naming is based on the following logic:
* self.binaries from "packages" files
(by this logic, it ought to be "self.packages", but that is
for later)
* The "_X_a" is short for "[<suite>][<parch>]" look ups.
* binaries_X_a and provides_X_a are the specialised parts of
packages_X_a that deal with (real) binary packages and
provides (i.e. virtual packages) respectively.
Signed-off-by: Niels Thykier <niels@thykier.net>
Extract a specialised iter_packages_hint from iter_packages that only
deals with applying hints. This simplifies iter_packages AND avoids
having to re-compute the uninstallability counters after each single
item in the hint.
This means that a hint can now avoid triggering expontential runtime
provided only that the "post-hint" stage does not trigger expontential
runtime. Previously the hint had to be ordered such that none of the
items in the hint caused such behaviour (if at all possible).
Signed-off-by: Niels Thykier <niels@thykier.net>
Avoid creating two dependency clauses for dependencies emulating a
"version range" a la:
Depends: pkg-a (>= 2), pkg-a (<< 3~)
Previously this would create two clauses a la:
- (pkg-a, 2, arch), (pkg-a, 3, arch)
- (pkg-a, 1, arch), (pkg-a, 2, arch)
However, it is plain to see that only (pkg-a, 2, arch) is a valid
solution and the other options are just noise. This patch makes
Britney merge these two claues into a single clause containing exactly
(pkg-a, 2, arch).
Signed-off-by: Niels Thykier <niels@thykier.net>
The "new" auto hinter relies on partial ordering to determine, when
what can migrate (and what needs to migrate at the same time). At the
same time, it leverages on "_compute_groups" to allow it to include
"removals" in its hints.
Signed-off-by: Niels Thykier <niels@thykier.net>
Avoid smooth-updating libraries in hints, when all of their reverse
dependencies will certainly disappear in the same hint.
Note that in "hint"-hint, reverse dependencies removed in the
following "full run" will not cause the smooth-updated library to be
removed. Instead these will still be removed in the end as usual, but
in some cases that is too late.
Signed-off-by: Niels Thykier <niels@thykier.net>
Rename find_upgraded_binaries into _compute_groups. The new method
will also compute what binaries will be updated in or added to testing
after migration.
Signed-off-by: Niels Thykier <niels@thykier.net>
Based on Colin Watson's code to do the same from the "britney2-ubuntu"
repository[1] revision 306, 308 and 309.
Notable differences include:
* output include version of source package being removed
* output prefix removals with a "-" (otherwise it would be identical to
a upgrade/new source with the change above).
[1] http://bazaar.launchpad.net/~ubuntu-release/britney/britney2-ubuntu/revision/306
Signed-off-by: Niels Thykier <niels@thykier.net>
This introduces a new variable HINTSDIR, which overrides the location of the
Hints dir (normally it is read from the UNSTABLE dir).
Please note that this is the location of the dir that contains the Hints dir.
Signed-off-by: Niels Thykier <niels@thykier.net>
This introduces a new variable OUTPUTDIR, which overrides the location where
the new dates file is written. This allow to run britney against a read-only
copy of the data.
Signed-off-by: Niels Thykier <niels@thykier.net>
Set the default to maxint until we've read something.
Reported-by: Ivo De Decker <ivo.dedecker@ugent.be>
Signed-off-by: Julien Cristau <jcristau@debian.org>
The new Installability Tester (IT) module replaces the remaining
C-parts. Unlike C-implementation, it does not give up and emit an
"AIEEE" half-way through.
In order to determine installability, it uses two sets "musts" and
"never". As the names suggest, the sets represents the packages that
must be (co-)installable with the package being tested and those that
can never be co-installable. For a package to be installable, "musts"
and "never" have remain disjoint.
These sets are also used to reduce the number of alternatives that are
available to satisfy a given dependency. When these sets are unable
to remove the choice completely, the new IT defers the choice to later.
This occasionally reduces backtracking as a later package may conflict
or unconditionally depend on one of the remaining alternatives.
Signed-off-by: Niels Thykier <niels@thykier.net>
We stopped populating the element with real data some time ago, it's
time to drop it entirely.
Signed-off-by: Adam D. Barratt <adam@adam-barratt.org.uk>
Multiarch adds a Depends: foo:any syntax, permitted only if the
target of the dependency is "Multi-Arch: allowed". This has
been supported by dpkg and apt for some time and is now safe to
use in unstable.
[Adam D. Barratt: adjusted to use consts.py]
Signed-off-by: Adam D. Barratt <adam@adam-barratt.org.uk>
As HintItem is now redundant, also replace it with a new class -
UnversionnedMigrationItem - and migrate users of the classes to use
the new versions.
Signed-off-by: Adam D. Barratt <adam@adam-barratt.org.uk>
Where possible, avoid creating a list only to discard immediately
afterwards. Example:
"""
for x in sorted([x for x in ...]):
...
"""
Creates a list, passes it to sorted, which generates a new list and
sorts that copy. Since sorted accepts an iterable, we can avoid the
"inner" list and just pass it a generator expression instead.
Signed-off-by: Niels Thykier <niels@thykier.net>
By moving the package loop inside register_reverses, it will be
invoked a lot less (reducing the overhead of invoking functions).
Signed-off-by: Niels Thykier <niels@thykier.net>
Beside some "minor differences" they were computing the same "tree"
(read: "graph"), so merge them into one (get_reverse_tree) and
properly document return value and special cases.
Signed-off-by: Niels Thykier <niels@thykier.net>
Rewrite the arguments of find_upgraded_binaries to not use an instance
of MigrationItem. We want to call it at a time where we have not
created MigrationItems yet.
Signed-off-by: Niels Thykier <niels@thykier.net>
A removal hint will generate both source and per-arch excuses if the
version of the source package differs between testing and unstable. If
the source versions are the same then only the per-arch excuses will
be generated.
Signed-off-by: Adam D. Barratt <adam@adam-barratt.org.uk>
In rare cases with hints with overlapping virtual packages provided by
different sources, this can make a difference.
Signed-off-by: Adam D. Barratt <adam@adam-barratt.org.uk>
If there are multiple versions of an arch:all package in unstable (due
to outdated or no longer built arch:any packages) then only one of them
should be recorded in the list of binary packages built from the source
package. Otherwise we may try and remove the binary package from various
lists multiple times, leading to crashes.
Signed-off-by: Adam D. Barratt <adam@adam-barratt.org.uk>
Given a source which provides two packages and has different versions
in testing and unstable, binNMUs in unstable corresponding to the
older source version should not be considered as migration candidates.
For example:
testing
-------
source 1
bin 1 arch1
bin 1 arch2
unstable
--------
source 2
bin 2 arch1
bin 1+b1 arch2
The binary migration on arch2 should not be considered a candidate.
Signed-off-by: Adam D. Barratt <adam@adam-barratt.org.uk>
Although this should never happen, rather than crashing if one of the
versions is none, simply indicate that they are unequal.
Signed-off-by: Adam D. Barratt <adam@adam-barratt.org.uk>
Although this isn't an issue during normal runs, the excuses might be
built multiple times during a hint-tester run and should not accumulate
during the run.
Signed-off-by: Adam D. Barratt <adam@adam-barratt.org.uk>
For those hints which don't cause an immediate run (i.e. other than
easy, hint and force-hint), re-build the excuses after adding the
hint so that the actions are accounted for in later hints.
Signed-off-by: Adam D. Barratt <adam@adam-barratt.org.uk>
A binNMU does not rebuild architecture:all packages. For migrations via
unstable this is not a problem as the packages corresponding to the
source upload are still present. However, for *pu migrations, the set of
packages considered only includes architecture-specific packages. In
order to avoid installability issues with packages in testing which
depend on the arch:all packages, we leave the existing arch:all packages
in testing and only consider the arch-specific packages for migration.
Signed-off-by: Adam D. Barratt <adam@adam-barratt.org.uk>
The test only needs to consider whether any binaries exist on a given
arch, not how many of them there are (or indeed which binaries they are)
Signed-off-by: Adam D. Barratt <adam@adam-barratt.org.uk>
When checking whether a tpu source has built on a particular arch, we
should only consider binaries produced by the latest version of the
source package in tpu.
Signed-off-by: Adam D. Barratt <adam@adam-barratt.org.uk>
Originally when binNMUs for packages in testing were scheduled, the
binaries would be installed into tpu with no accompanying source. This
allowed the "removed binary" portions of should_upgrade_srcarch() to be
skipped (as britney had generated a faux source record).
dak now adds the source package to tpu in such cases which lead to the
"removed binary" checks being applied to binNMUs in tpu with potentially
destructive consequences. For example, if a package with amd64 and i386
binaries in testing were binNMUed on just amd64, britney would notice
that there were no i386 binaries in tpu and subsequently remove the i386
binaries from testing as well.
In order to resolve this, we skip the check for removed binaries when
building excuses for a binary-only migration via *pu.
Signed-off-by: Adam D. Barratt <adam@adam-barratt.org.uk>
The primary difference between the parsing / output of excuses for *pu
and unstable unblocks is the messages displayed. We can therefore remove
some duplication by having the same code handle both, outputting the
appropriate message.
Where a *pu package is also the subject of a "block" (most likely during
a freeze) we only supply the "needs approval" or "approved" message;
previously both "needs approval" and "not touching due to block" were
output, which is redundant. We ensure that there is always a dummy
"block" hint for *pu packages to provide the "needs approval" behaviour.
Signed-off-by: Adam D. Barratt <adam@adam-barratt.org.uk>
An "approve" hint is effectively an unblock for tpu packages and britney
is already quiite happy to parse "unblock $pkg/$tpuversion".
We allow the old name to be used for compatibility and replace it with
"unblock" internally.
Signed-off-by: Adam D. Barratt <adam@adam-barratt.org.uk>
A dependency on an arch-specific package which is not a valid candidate
should lead to the depending package not being a candidate.
For now we ensure that the generated excuses output remains the same,
so that we don't have to wait for consumers to adapt to a new format.
Changing the output format should be revisited at a later point.
See Debian bug #693068.
Signed-off-by: Adam D. Barratt <adam@adam-barratt.org.uk>
The code using the variables was refactored in 694d614b. As a result
they were still set in iter_packages() but never subsequently used.
Signed-off-by: Adam D. Barratt <adam@adam-barratt.org.uk>
Previously a package which became obsolete during a run would not be
automatically removed until the next run. This was due to the fact that
sources[][BINARIES] is not updated during the run. Instead, we build a
list of source packages which produce at least one binary and then
remove any packages not in that list.
Signed-off-by: Adam D. Barratt <adam@adam-barratt.org.uk>
"not force and not earlyabort" simplifies to "not earlyabort" rather
than "not force", as an easy hint would set "earlyabort" but not
"force".
Signed-off-by: Adam D. Barratt <adam@adam-barratt.org.uk>
All callers of get_reverse_tree compute the same modification of its
return value, so move that computation into get_reverse_tree.
Signed-off-by: Niels Thykier <niels@thykier.net>
When processing a hint of the form "easy pkgX libX" where libX would be
a candidate for smooth updates because pkgX/testing depends on it but
pkgX/unstable does not, and there are no other reverse dependencies,
the old binary from libX can simply be dropped straight away.
Signed-off-by: Adam D. Barratt <adam@adam-barratt.org.uk>
The only test currently implemented is to ensure that any prospective
hint contains at least one item beyond the hint name. This prevents
lines in a hint file consisting simply of e.g. "easy" being added to
the hint list and causing later processing to abort with an error.
Signed-off-by: Adam D. Barratt <adam@adam-barratt.org.uk>
This causes Multi-arch dependencies like "pkg:i386" to show up as
unsatisfiable in excuses.
Previously, the dependency would be checked on the wrong architecture
(if available) and cause the package to become a valid candidate. The
package would still be prevent from migrating as the installability
checker does not know of the "pkg:i386" package.
Signed-off-by: Niels Thykier <niels@thykier.net>
Niels Thykier
Paul Wise
Emilio Pozuelo Monfort
Adam D. Barratt
Ivo De Decker
Cyril Brulebois
Anthony Towns
Julien Cristau
Ivo De Decker
Colin Watson
Colin Watson