Time Nick Message 04:51 sm owst: it redirects properly after merging now 09:23 gal_bolle hi owst 09:24 owst hi gal_bolle 13:25 kowey hello! 13:38 owst Afternoon kowey 15:55 kowey hi bsrk, ready to start? 15:55 bsrk yes 15:55 kowey how did your week go? 15:56 bsrk My school semester just started, 15:56 bsrk an unfortunately the internet facility does not support ssh connections. I could not upload my work into the repo. 15:57 kowey hmm 15:57 bsrk I will upload it tomorrow using a third party connection. 15:57 kowey are you going to be able to continue working in a full time capacity? 15:57 bsrk yes. 15:58 bsrk my classes are like 18hrs per week 15:58 bsrk and my gsoc work is like 20hrs per week. 15:58 kowey you do realise that GSoC runs on 40hrs per week? 15:59 kowey I mean, nobody is here to follow you around with a stopwatch… 15:59 bsrk I am talking about how *I* did it. But yes, it is supposed to be 40 hrs a week. :-) 16:00 kowey hmm 16:01 bsrk So, this week I wrote the tests planned 16:01 bsrk (1) Testing based on properties of patch index structures 16:02 bsrk (2) Expand the shell test, and testing with expected correlation 16:02 bsrk I have completed the timing tests 16:02 bsrk and updated the google docs 16:03 kowey less computer crashing this time 16:03 bsrk I have updated the documentation on structures. 16:03 bsrk Actually they continued to crash 16:03 bsrk I have split up the work 16:04 bsrk into smaller chunks 16:04 bsrk and ran them indivdually 16:05 kowey have you learned anything from the timing tests? 16:07 bsrk Patch index is worth it. :-) 16:07 bsrk It saves 15sec on average at cold in annotate, and 13 sec on average at cold in changes. 16:08 kowey what is average? 16:09 kowey hmm, sorry not a clear question 16:09 kowey have you thought carefully about what you mean when you say “on average” 16:09 kowey is that notion even sensible? 16:09 bsrk https://docs.google.com/spreadsheet/ccc?key=0AtdxF5AhJcuwdHFlMm0xd1poeTROYW1KN0ZGR2U0enc 16:09 bsrk https://docs.google.com/spreadsheet/ccc?key=0AtdxF5AhJcuwdGdrT2lLMGZXbktZblFxWmhxNkxXRVE 16:09 kowey are you familiar with the distinction between the mean and the median? 16:10 bsrk Of course. In this case I took the mean of all the files in darcs repo. 16:10 kowey (I'm not sure how rigorous we really need to be about this) 16:10 kowey I guess what I'm concerned about is “how meaningful a metric is this?” 16:11 kowey like what's the distribution of times and savings? 16:11 kowey well, it's hard for me personally to expect much more statistical/experimental sophistication when I myself don't know what I'm doing 16:12 bsrk For what it is worth, it seems that patch index saves more time than costs. 16:12 bsrk My experience is only with hot repos, 16:12 bsrk but I am not noticing any lag for update 16:12 bsrk and the lag of create is not too bad. 16:13 bsrk Most of my queries with changes, annotate give instantaneous responses. 16:14 kowey have you had time to look into darcs-benchmark? 16:14 bsrk I did not look into it any further. :-( 16:14 kowey not enough time? 16:15 kowey sorry, perhaps a bit harsh 16:15 bsrk Is it a priority? I want to finish up the formal goals first. 16:15 kowey tell me about the mailing list discussion 16:15 bsrk I don't really have an excuse of "not enough time". :-) 16:15 kowey how did that go? 16:16 bsrk There were a few good ideas. But mostly I am concerned with the amount of miscommunication. 16:17 kowey what do you mean? 16:18 bsrk 1) Feedback at patch index creation: If you run patch index on an old repo, you will get a message saying patch index is being created. 16:20 bsrk 2) Patch index is updated/created, after the actual operations. ie you can do a control-c, and the operation will still succeed with an old pi state. 16:22 bsrk 3) If patch index is created at get, it will be fast. Even if it is created from a cold state, it is still creatively not bad (around 4x the average changes) 16:23 bsrk So, basically the good suggestions were: 16:24 bsrk 1) Renaming no-patch-index to disable-patch-index. 16:25 bsrk 2) An explicit statement that ctrl-c is okay when pi is being created/updated. The user should know that the main operation is over though.. 16:28 kowey what are you going to do now that people have given you feedback on the list? 16:29 bsrk I will clear up issues with a follow up post, and implement the suggestions. :-) 16:30 kowey so anything else to report for the week? 16:31 bsrk I basically completed goal 4 of testing. 16:32 bsrk I asked Benedikt for review of goal 2 (using pi in changes/annotate) 16:33 bsrk and goal 3 is in progress. (I started a little, and hope to make more progress in the weekend) 16:34 kowey have you been meeting with Benedikt these few weeks? 16:35 bsrk yes? 16:36 bsrk yes. :-) 16:36 kowey how's that going? 16:37 bsrk We discuss some issues, and he has a few suggestions. 16:41 kowey ok 16:41 kowey so weeks 11-13 16:41 kowey what have you got planned? 16:41 kowey you mentioned getting back to the people who took the time to respond to your request 16:43 bsrk I will finish up documentation, Implement suggestions from Benedikt's reviews, and respond back to the users post. (next week) 16:45 kowey there's also darcs-benchmark you could work on 16:45 kowey but I do understand that documentation can be pretty hard work 16:45 kowey as usual, I'll do what I can to give you feedback 16:45 kowey anything else? 16:46 bsrk yes. I will work on benchmark after I finish up the documentation on tests. 16:46 kowey well, I guess I have some stuff to report 16:46 kowey I'm still working on rebasing your patches 16:47 kowey it's slow going, partly because I'm being a bit perfectionistic 16:47 kowey trying to tease things about into patches that would be easy to review 16:47 * Heffalump is having the same issue with rebasing rebase :-) 16:47 kowey :-) 16:47 Heffalump though I'm kind of coming to the conclusion that most of the development history is uninteresting 16:47 Heffalump (I've also been fixing new problems I found in rebase as I go) 16:47 kowey I am getting the hang of using the rebase stack 16:47 kowey for example, if I want to split a patch over its precedessors 16:48 kowey I now, unrecord, amend, suspend, amend, suspend, amend, unsuspend, unsupsend 16:48 kowey probably a bit more fiddly than git users would like 16:48 kowey but being able to focus on only one thing at a time like that with the darcs interactive ui is also nice 16:48 kowey (not fiddly like complicated, just that you have to mentally work out how to achieve what you want) 16:48 kowey oh sorry, bsrk, I'm letting myself get distracted 16:49 kowey so yeah, basically still rebasing your work 16:49 kowey otherwise 16:50 kowey I'll have to admit I'm a bit unhappy about this 20h work week thing 16:50 kowey there's not much I can say here, now, because it's just an immediate reaction 16:50 kowey but I'll have to take some time to sort through my feelings/thoughts about it 16:51 kowey and figure out a response that would be good for you and past/future GSoCers 16:51 kowey sorry if I'm being a bit too upfront about it 16:51 bsrk Nah, I understand. 16:52 kowey alright, well good luck for next week 16:52 kowey do use me for documentation feedback 16:52 kowey I'm mostly offline Mon-Wed as I'm focusing on work 16:53 kowey see you around 20:27 davidsarah hello 20:27 davidsarah we have a reproducible test case where 'darcs get --tag' does the wrong thing, in darcs 2.5.2 20:27 davidsarah (am installing 2.8.1 now to test with that) 20:27 davidsarah does that sound like a known bug? 20:28 davidsarah it seems similar to this: http://irclog.perlgeek.de/darcs/2012-05-28 20:30 davidsarah script that reproduces the bug: http://codepad.org/vH2N04lo 20:31 davidsarah expected output (using darcs 2.4.4): http://codepad.org/OUae28Km 20:32 davidsarah incorrect output (using darcs 2.5.2): http://codepad.org/U5MZX830 20:33 davidsarah same problem with 2.8.1 20:34 davidsarah the problem seems to be that the later darcs versions are not unapplying patches that need to be unapplied in order to reproduce the tagged revision 20:53 Heffalump davidsarah: doesyour tag situation sound similar? 20:53 Heffalump i.e. could the tag be "not clean", i.e. have patches before it that aren't actually in the tag 20:54 Heffalump I thought I actually made a test case for dmwit's problem 20:54 Heffalump but I'd have to check the bug tracker as I can't see any mention of it in the onversation log 21:25 * davidsarah reads the log again 21:27 davidsarah our test case seems simpler than that one, but it *could* be the same bug 21:28 davidsarah yes, the "Unapplying 0 patches" (when you'd expect some patches to be unapplied to get to the tag) sounds the same 21:30 Heffalump ah, I did flag a bug about it: http://bugs.darcs.net/issue2199 21:30 Heffalump no test case, but it suggests I did reproduce it as that 21:31 davidsarah I'm unclear what you mean by a tag being "dirty" 21:32 Heffalump "not clean" as above - an internal representation thing really, but it bleeds into darcs beaviour sometimes 21:32 davidsarah please look at http://codepad.org/vH2N04lo and tell me what it is doing to produce a dirty tag, if that's the problem 21:33 Heffalump ooh, a test case, great! :-0 21:33 davidsarah does "dirty" mean that some patches have to be unapplied to get to the tag? 21:33 Heffalump and it fails in mut? 21:34 Heffalump no, it means that some patches that are before the tag in the repo can be unpulled without unpulling the tag 21:34 davidsarah hmm 21:34 Heffalump assuming I'm right that it fails in mut and not in orig, the reason is the record you do in mutant before pullnig th tag 21:34 davidsarah (21:31:09) davidsarah: expected output (using darcs 2.4.4): http://codepad.org/OUae28Km 21:34 davidsarah (21:32:29) davidsarah: incorrect output (using darcs 2.5.2): http://codepad.org/U5MZX830 21:34 davidsarah (21:33:12) davidsarah: same problem with 2.8.1 21:35 Heffalump it's not your fault, it's a clear bu 21:35 * davidsarah nods 21:35 Heffalump and yes, you're getting exactly what I'd expect if that ws the characterisation of the bug. 21:35 davidsarah ok, thanks 21:35 Heffalump I guess I should fix it instead of talking about it so much :-) 21:35 davidsarah :-) 21:36 davidsarah (22:34:57) Heffalump: assuming I'm right that it fails in mut and not in orig, the reason is the record you do in mutant before pulling the tag 21:37 davidsarah yes, that's what I thought it must be