Perl 6 - the future is here, just unevenly distributed

IRC log for #darcs, 2014-11-16

| Channels | #darcs index | Today | | Search | Google Search | Plain-Text | summary

All times shown according to UTC.

Time Nick Message
00:23 alexei joined #darcs
00:35 amgarchIn9 joined #darcs
02:47 ilbot3 joined #darcs
02:47 Topic for #darcs is now http://darcs.net/ | logs: http://irclog.perlgeek.de/darcs/ | darcs 2.8.5 is out http://darcs.net/Releases/2.8
04:11 mizu_no_oto joined #darcs
08:16 lelit joined #darcs
10:26 amgarchIn9 joined #darcs
11:52 amgarchIn9 joined #darcs
13:07 amgarchIn9 joined #darcs
13:39 mizu_no_oto joined #darcs
13:45 mizu_no_oto joined #darcs
15:39 burp joined #darcs
16:44 mizu_no_oto joined #darcs
17:03 gh_ joined #darcs
17:27 amgarchIn9 joined #darcs
18:07 gh_ Heffalump, I have these two patches but I'm not sure to push them to screened as they may diverge too much from yours.. they actually make patch index *use* slighly slower for annotate, but I believe the code is more understandable: http://hub.darcs.net/gh/darcs-screened_1
18:11 gh_ the first patch directly concerns annotate: the old algorithm has an advantage when you annotate a file whose patches are only at the beginning of the repository, with a lot of patches afterwards. While the new algorithm has to scan the whole history always.
18:16 gh_ the second patch removes all indexing when storing PI on disk. this makes the code simpler and on-disk size bigger. but also is changes the infomap (file -> patch) into a bloom filter, that is we only store the first Word32 of patch hashes and tolerate false positives. this enables this particular file to come back to its original size, and has the promise of being faster to look into.
18:33 nebnesknarf joined #darcs
19:43 Heffalump hi nebnesknarf, sorry about this encoding mess..
19:44 Heffalump gh_: go ahead and push anything of yours that you're happy with
19:44 Heffalump mine are all conceptual experiments really, I can easily redo them if they still make sense
19:44 Heffalump nebnesknarf: I think you're right that we should be applying decodeString before parsing
19:45 Heffalump gh_: and what you describe sounds sensible to me
19:58 lelix joined #darcs
20:04 Heffalump gh_: does it still have the linear memory usage? i.e. should I keep trying to sort that out? I think my approach is probably the right one, just needs to be implemented more carefully :-)
20:06 gh_ Heffalump, I haven't tried it on top of your patches, so I don't know but there is no change in the code that creates the PI
20:10 Heffalump ok
21:16 gh_ joined #darcs
21:26 Rastus_Vernon joined #darcs
21:35 nebnesknarf Hi, i was offline, thanks for getting back to me
21:35 nebnesknarf I am not sure I know what to do but the last patch I attached to the tracker looks promising
21:36 nebnesknarf BTW, who is who here?
21:37 nebnesknarf Heffalump: you must be ganesh, says my chat program
21:37 nebnesknarf Ah, so gh_ is Guillaume. Git it now
21:37 gh_ hi :D
21:37 gh_ correct
21:39 nebnesknarf Heffalump: could you still take a look at what I have sent?
21:39 nebnesknarf It is based on you changes, which I like.
21:40 nebnesknarf I could use a bit of help with the last failing test (workingdir.sh). Something is still wrong with the --to-patch option and I can't seem to figure out what it is.
21:42 nebnesknarf Last time I checked, it looked as if darcs get --to-patch no longer works.
21:46 Heffalump nebnesknarf: I'll take a look (might be a day or two before I have time properly)
21:46 Heffalump I'm sure I used darcs get --to-patch a few days ago, but that might have been using 2.8, the darcs I pick up from my path can be a bit random
21:48 amgarchIn9 joined #darcs
21:58 mizu_no_oto joined #darcs
21:59 nebnesknarf Heffalump: this happens is with your patch that normalizes the flag list; I am not 100% it did that before, but think it did not.
21:59 nebnesknarf OMG my English is bad today I should go to bed
22:02 Heffalump ah, hmm. That's unfortunate, I thought that patch was about making sure we had the defaulting behaviour from before the new options system even when using [DarcsFlag],but it seems to be getting caught out
22:06 Heffalump on the other hand as far as I know, workingdir.sh did work just with my patch - the utf8 failure was the only one I ran into.
22:06 Heffalump anyway, will check again carefully
22:07 nebnesknarf Yes, me too, again.
22:08 nebnesknarf Also, Owen's patches are nice and would like to screen them.
22:08 nebnesknarf But they conflict with your patch. WHich one should I amned?
22:09 nebnesknarf I am minded to push Owen's changes first, then ours amended. That ok?
22:12 nebnesknarf Ah, now that I think about it: the match flags are handled a bit differently from normal options.
22:13 nebnesknarf See D.UI.O.Matching
22:20 Heffalump we can always just resolve conflicts rather than amending/rebasing
22:21 Heffalump it's only when the conflicts get out of hand that it becomes really necessary to do that
22:23 nebnesknarf So I apply with --mark-conflicts, then record a resolution, then push?
22:23 nebnesknarf Could as well rebase apply, then unsuspend and amend.
22:24 nebnesknarf I am feeling more confident with this workflow. Are you okay with that?
22:25 Heffalump the former keeps the identity of the original patches, which makes it easier to collaborate on them
22:26 Heffalump so I prefer the former, but I'm not completely wedded to it. I did actually try to resolve conflicts first with your options patch, but the accumulation was too much so I ended up having to rebase.
22:27 nebnesknarf OK. I see you have screened things. Good.
22:28 nebnesknarf Now, that is strange. I made a fresh clone of screened a few minutes ago and after I compile it the tests all run through. How can that be???
22:28 Heffalump I haven't screened the options type as a parameter thing
22:28 Heffalump what were you referring to when you said I have screened things?
22:29 nebnesknarf Oops, my mistake, I had a wrong upstream.
22:29 nebnesknarf Sorry for noise.
22:30 nebnesknarf Still: when I compile the current screened I no longer get any failures, not even in utf8.sh.
22:30 Heffalump isn't that to be expected? We didn't push any patches to screened that would break it.
22:31 Heffalump your removal of decodeString isn't in screened, and neither is my options type as a parameter
22:31 Heffalump and your convert.sh fix is in screened, so the shell tests should be clean (the unit tests are another matter as you've been discovering, but those have been a bit of a mess for a long time)
22:32 nebnesknarf (yes I ran with unit=no)
22:32 nebnesknarf Ooookay. So I have been confused about the source of the problem.
22:33 nebnesknarf I will now pull your and my patches one at a time and each time run the tests and see what happens.
22:35 Heffalump what I know is this: my options-type-as-a-parameter breaks utf8.sh because of the parse-unparse roundtrip problem
22:35 Heffalump your decodeString removal breaks utf8.sh for reasons I don't know the precise details of but it's not that surprising
22:35 Heffalump and I haven't looked at any of the recent patches to see what they break if anything.
22:36 Heffalump and don't go near the patches I've been sending to do with the patch index :-)
22:41 nebnesknarf I have now pull your options-type-as-a-parameter patch and I get three failing tests: rebase-unsuspend-to-patch.sh, utf8.sh, and workingdir.sh, just as before.
22:43 nebnesknarf (I wouldn't dare go near things like patch index ;-)
22:44 Heffalump utf8.sh is great at finding problems, but it's a pain in the backside to debug (and as you realised, it gets skipped too often)
22:46 nebnesknarf ...and doesn't work for me at all unless I add a --skip-long-comment somewhere at the beginning.
22:46 Heffalump oh dear..
22:46 Heffalump that must be something to do with editors, but the test suite is supposed to set those properly
22:47 mizu_no_oto joined #darcs
22:47 nebnesknarf It's in darcs record -i < interactive_whatever; I guess it has to do with changing the default from prompt to always open editor
22:47 Heffalump but it does pass for me with screened without changing it
22:49 nebnesknarf I have DARCS_EDITOR and EDITOR set and I have record prompt-long-comment in my defaults
22:50 Heffalump the test harness is supposed to override all those
22:50 Heffalump it sets up a private home directory to run them in
22:50 Heffalump oh, maybe it doesn't actually override EDITOR and DARCS_EDITOR
22:50 Heffalump that seems like a bug
22:51 nebnesknarf Good, so this can be fixed.
22:52 nebnesknarf BTW, another problem I have is with setting 'changes interactive' in defaults
22:52 Heffalump I'm getting quite a few failures - more than the ones you mentioned - with your latest bundle in issue1226
22:52 Heffalump patch1226 that is
22:53 nebnesknarf The one with _and_4_others at the end?
22:54 Heffalump send, workingdir, send-encoding, issue381, issue2204-send-mail, issue1344_abort_early_cant_send
22:54 Heffalump yes, that. I guess 'send' is the common factor in most of the extra failures I get.
22:54 nebnesknarf ??? I have never had failures in one of those. Perhaps these tests are also disabled for me for some reason?
22:55 Heffalump quite possible...if you run them, you should see 'Skipped' if they are
22:56 Heffalump I'm just retrying with vanilla screened to check it's not some new problem on my machine
22:57 nebnesknarf And I'm doing cabal clean and will rebuild
22:58 Heffalump ok, it does seem to be coming from those patches. But I really need to go to bed now.
23:02 nebnesknarf Yes, me too. Thanks for taking time to check this.
23:05 nebnesknarf BTW, still the same result here, too: one test failing (workingdir)
23:05 nebnesknarf Good night
23:05 nebnesknarf left #darcs

| Channels | #darcs index | Today | | Search | Google Search | Plain-Text | summary