1 2011-09-05T00:33:34 *** raignarok
2 2011-09-05T02:59:21 *** grzywacz
3 2011-09-05T07:59:17 <dreimark> moin
4 2011-09-05T08:23:07 *** pkumar
5 2011-09-05T08:25:28 <pkumar> moin
6 2011-09-05T08:32:15 <dreimark> pkumar: read backlog before 18:28 may be you can help there
7 2011-09-05T08:35:16 *** pkumar
8 2011-09-05T09:37:16 *** raignarok
9 2011-09-05T11:23:54 *** CIA-23
10 2011-09-05T11:23:59 *** CIA-65
11 2011-09-05T11:38:02 *** greg_f
12 2011-09-05T12:33:16 *** mkerrin
13 2011-09-05T16:16:05 <ThomasWaldmann> moin
14 2011-09-05T16:19:51 *** pkumar
15 2011-09-05T16:21:04 <pkumar> dreimark : hey, sure, looking into it.
16 2011-09-05T16:47:41 *** brunomartin
17 2011-09-05T16:57:20 <ThomasWaldmann> hi brunomartin
18 2011-09-05T16:57:46 <brunomartin> hi Thomas!
19 2011-09-05T17:04:14 <ThomasWaldmann> brunomartin: see query
20 2011-09-05T17:30:23 <pkumar> ThomasWaldmann : hi,
21 2011-09-05T17:30:46 <pkumar> I was looking at those tests in hg
22 2011-09-05T17:31:04 <pkumar> test_backends_hg*
23 2011-09-05T17:32:14 <ThomasWaldmann> pkumar: yes? did you find anything?
24 2011-09-05T17:32:36 <ThomasWaldmann> could you reproduce the issue?
25 2011-09-05T17:32:51 <pkumar> yeah I reproduced it
26 2011-09-05T17:33:49 <pkumar> seems like once a test is marked xfail, it stays like that until the py.test run stops. currently all the tests share test_backend
27 2011-09-05T17:34:15 <pkumar> i.e. tests in test_backends
28 2011-09-05T17:35:06 <pkumar> so, when the test items are run again after it is marked xfail due to test_backend_hg
29 2011-09-05T17:35:40 <pkumar> it shows xfail, though other tests in
30 2011-09-05T17:36:17 <pkumar> not in test_backends run nicely without xfail
31 2011-09-05T17:36:47 <pkumar> so, one solution is to mark xfail the test functions in hg_backeds individually
32 2011-09-05T17:37:24 <pkumar> have separate tests for test_hg_backends
33 2011-09-05T17:37:41 <pkumar> ThomasWaldmann : ^
34 2011-09-05T17:39:40 <pkumar> ThomasWaldmann : what do you suggest?
35 2011-09-05T17:41:55 <pkumar> test_backend_hg*
36 2011-09-05T17:43:22 <ThomasWaldmann> hmm, maybe it should not mark the class as xfail, but rather the instance?
37 2011-09-05T17:43:57 <pkumar> yeah, that should do I guess
38 2011-09-05T17:44:38 <ThomasWaldmann> btw, I guess all these tests need quite some more work in future
39 2011-09-05T17:45:58 <ThomasWaldmann> at the weekend i was trying (once again) to do some backend work (like getting rid of item metadata for example), but such stuff always gets way to complicated somehow
40 2011-09-05T17:46:33 <pkumar> yeah, I've got an habit of reading backlogs ;)
41 2011-09-05T17:46:35 <ThomasWaldmann> partly due to tests, partly due to twisted things in the moin code
42 2011-09-05T17:47:08 <ThomasWaldmann> so, I am not yet sure, but I am thinking about trying to rewrite the storage tests
43 2011-09-05T17:47:25 <pkumar> all of them?
44 2011-09-05T17:48:09 <ThomasWaldmann> one can maybe reuse some code, but somehow those tests are much too complicated and might not match the current vision of storage
45 2011-09-05T17:49:05 <ThomasWaldmann> (current vision is that storage is rather simple and lots of stuff is done at index level, not at storage level)
46 2011-09-05T17:49:15 <ThomasWaldmann> for example:
47 2011-09-05T17:49:52 <ThomasWaldmann> initially we had the idea that storage code provides a history() functionality, yielding all changes in sorted order
48 2011-09-05T17:50:23 <ThomasWaldmann> whoosh can do that in a very easy way and likely also much more efficiently using the index
49 2011-09-05T17:50:40 <ThomasWaldmann> so, we don't need history in storage implementations any more
50 2011-09-05T17:50:59 <pkumar> hmm... it's been removed right?
51 2011-09-05T17:51:42 <pkumar> history stuff?
52 2011-09-05T17:53:29 <ThomasWaldmann> yes, i have removed that already
53 2011-09-05T17:53:35 *** mkerrin
54 2011-09-05T17:53:38 <ThomasWaldmann> but there is lots more to simplify
55 2011-09-05T17:53:56 <ThomasWaldmann> e.g. has_item - we can do that very quickly via index, too
56 2011-09-05T17:54:10 <ThomasWaldmann> but, if one tries this, lots of tests are breaking
57 2011-09-05T17:54:37 <ThomasWaldmann> likely just because they are testing on wrong level right now
58 2011-09-05T17:54:59 <ThomasWaldmann> e.g. the fs2 tests test if fs2's has_item works correctly (low level)
59 2011-09-05T17:55:56 <ThomasWaldmann> if I want to use indexed has_item, it needs the layered storage set up, so it is router middleware with indexing mixin -> fs2 backend and tests has_item on router level then
60 2011-09-05T17:56:34 *** raignarok
61 2011-09-05T17:56:56 <ThomasWaldmann> so some (all?) tests would run on top of router/indexing -> backend X, for X in all available backends
62 2011-09-05T18:01:51 <pkumar> yeah, seems like some changes in the tests.
63 2011-09-05T18:02:03 <pkumar> are required in that case
64 2011-09-05T18:06:20 <pkumar> current test architecture may still work, like using the common test test_backends and just change the individual tests to work likewise
65 2011-09-05T18:11:14 *** raignarok
66 2011-09-05T18:19:28 *** raignarok
67 2011-09-05T18:30:49 <dreimark> bbl
68 2011-09-05T19:09:27 *** greg_f
69 2011-09-05T19:45:40 *** pkumar
70 2011-09-05T22:38:52 *** raignarok
71 2011-09-05T22:47:46 <dreimark> re
72 2011-09-05T22:59:30 * ThomasWaldmann just booked his flights \o/
73 2011-09-05T23:05:26 <dreimark> ThomasWaldmann: how long do you stay in the US ?
74 2011-09-05T23:13:43 <ThomasWaldmann> see wiki, 1w
75 2011-09-05T23:14:15 <ThomasWaldmann> but i will stay in america for another wk
76 2011-09-05T23:57:25 <dreimark> seen
