Re: Bug#645558: Improve britney2 auto-hinter
On 2011-10-17 21:11, Joachim Breitner wrote:
> Hi again,
>
> Am Montag, den 17.10.2011, 20:34 +0200 schrieb Joachim Breitner:
>> Am Montag, den 17.10.2011, 20:09 +0200 schrieb Mehdi Dogguy:
>>> On 10/17/2011 05:46 PM, Joachim Breitner wrote:
>>>> Hi,
>>>>
>>>> Am Montag, den 17.10.2011, 10:45 +0200 schrieb Niels Thykier:
>>>>>> [1] http://release.debian.org/~nthykier/britney-tests
>>>>
>>>> can you make that git repository pullable via http or available
>>>> somewhere public (e.g. git.debian.org)?
>>>>
>>>
>>> You should be able to clone via http now.
>>>
>>> git clone http://release.debian.org/~nthykier/britney-tests.git
>>
>> Great, thanks. I bootstrapped it already via http using lftp and
>> "mirror", but for pulling changes this is much better.
>>
>> SAT-Britney passed the "basic" test just fine, now I’m adjusting the
>> test runners.
>
> thanks for the testsuite, Niels, it is great to have that.
>
You are welcome :)
> SAT-Britney succeeds in 10 out of 12 (one after adding a
> one-line-change, another one after fixing (IMHO) the test data). The
> remaining one are
> Running tree-circle-dependencies-huge-graph-no-hint... FAILED (0.666s)
> Running tree-circle-root-dependencies-no-hint... FAILED (0.011s)
>
> I think in the first case it’s because SAT-Britney finds the transition
> you are hinting at in these files, while the "expected" file expectes
> current britney’s behaviour which does not find the file:
> $ grep -2 -v '^[ ]' tree-circle-dependencies-huge-graph-no-hint/diff
> Added source packages:
> src-0-0 1.0-2
> src-0-1 1.0-2
> --
> src-9-8 1.0-2
> src-9-9 1.0-2
> Removed source packages:
> src-0 1.0-1
> src-1 1.0-1
> --
> src-9 1.0-1
> src-root 1.0-1
> Added binary packages:
> src-0-0 1.0-2 i386
> src-0-1 1.0-2 i386
> --
> src-9-8 1.0-2 i386
> src-9-9 1.0-2 i386
> Removed binary packages:
> pkg-0 1.0-1 i386
> pkg-1 1.0-1 i386
>
Actually this is a copy-waste error (assuming you are looking at the one
in t/). That expected file should not have been included as
hooks/post-setup generates a new one. Also to my knowledge, the current
britney implementation fails this test without the patch.
> In the second case, SAT-Britney seems to lack:
> $ grep -2 -v '^[ ]' tree-circle-root-dependencies-no-hint/diff
> Added source packages:
> leaf-a1 1.0-1
> Removed source packages:
> leaf-a1 1.0-2
> Added binary packages:
> leaf-a1 1.0-1 i386
> Removed binary packages:
> leaf-a1 1.0-2 i386
> I’ll see into that right away.
>
> Note that SAT-Britney ignores any "easy" hints, e.g. migration-chain
> works just fine without any manual intervention.
>
That hint was a copy-waste error and so are the "1.0-2" versions in
testing (as you noticed) :P
> I have put my repo here:
> http://anonscm.debian.org/gitweb/?p=users/nomeata/britney-tests.git;a=shortlog;h=refs/heads/master
> including a patch that makes runtests independent of the current working
> directory and that writes the diff between expected and real result to
> $testdir/diff.
>
>
> Greetings,
> Joachim
>
Thanks, I have merged the changes in your branch and added a couple of
other changes (particularly two new tests).
If you have any ideas for what the test-suite should be able to handle
(or how it should work), feel free to follow up on [1].
~Niels
[1] http://lists.debian.org/debian-release/2011/10/msg00178.html
Reply to: