Hi,
after having stopped working on SAT-Britney for about half a year, I
recently got a motivating mail and finally sat down and fixed a few of
the remaining annoyances to bring it into a usable form. It runs now
twice a day on ries and, thanks to Phil, the resulting hints are copied
to franck and executed by britney:
http://release.debian.org/britney/hints/satbritney
It only generates easy hints, and britney is configured to only allow
easy hints from it, so this should be pretty safe.
As pointed out by Julien, britney2 does not accept the currently
generated hints. This is due to a remaining difference in what
SAT-Britney and britney2 thinks is a valid transition: SAT-Britney will
allow the so-called “smooth upgrades” with every package, while britney
will only consider packages from libs and oldlibs. Another possible
cause for diverging behavior might involve package removals, we’ll see
if it becomes relevant in practice. I hope you can bear the occasional
false hint.
Another reason for strange hints could be the delay: SAT-Britney uses
the result of one britney run to generate the hints from, but these
hints are then processed with the next britney run, when some packages
have will have changed in unstable.
Besides that, SAT-Britney should be able to reliably find the largest
admissible transition, no matter how many packages need to migrate
together. My motivating example for that would be the next Haskell
migration with >400 source packages – once everything is built and
working in unstable, SAT-Britney should without further manual
interaction provide us with one huge and correct hint.
It passes most of the test cases in the britney test suite, including
some that britney2 fails (tree-circle-dependencies-huge-graph-new-young,
tree-circle-dependencies-huge-graph-no-hint,
tree-dependencies-no-hint-binnmu), the remainig cases involve force
hints (which SAT-Britney ignores), package renames and the generated
Source file (which is irrelevant for generating hints), see
http://git.nomeata.de/?p=sat-britney.git;a=blob;f=README.txt;hb=refs/heads/full-dependencies
for an up-to-date list of the failing test cases.
If there are any questions or suggests about SAT-Britney (besides „stop
it we don’t want it“, if possible), I’m happy to answer them.
Greetings,
Joachim
--
Joachim "nomeata" Breitner
Debian Developer
nomeata@debian.org | ICQ# 74513189 | GPG-Keyid: 4743206C
JID: nomeata@joachim-breitner.de | http://people.debian.org/~nomeata
Attachment:
signature.asc
Description: This is a digitally signed message part