Perl 6 - the future is here, just unevenly distributed

IRC log for #darcs, 2015-07-24

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

All times shown according to UTC.

Time Nick Message
01:30 mizu_no_oto joined #darcs
01:57 byorgey joined #darcs
01:57 byorgey joined #darcs
05:45 Heffalump no another todo (not urgent) - make the new tests run out of the box
06:28 alexei_ joined #darcs
09:44 alexei___ joined #darcs
10:35 mizu_no_oto joined #darcs
10:43 IbnFirnas joined #darcs
11:03 mizu_no_oto joined #darcs
11:09 ggherdov` joined #darcs
11:33 carter_cloud joined #darcs
12:05 pointfree joined #darcs
16:21 notdan ok, I managed to rebase the history further
16:21 notdan should simplify some things
16:21 notdan removed a bunch of patches that contained dead code
16:24 Pastaf joined #darcs
16:32 notdan Heffalump: did you manage to figure out the problem that you had with record?
16:41 Heffalump notdan: no, haven't investigated that further yet
16:42 Heffalump was actually just trying to build on Windows so I can test there - darcsden is always a bit of a pain on Windows especially if I haven't done it for a bit
16:42 Pastaf joined #darcs
16:43 notdan When you have a minute can you send me your 1) ~/.darcsden.conf 2) tree output 3) the repository in question
16:44 Heffalump tree output?
16:44 Heffalump the repo was just the darcsden-cli repo itself
16:45 Heffalump sent you darcsden.conf
16:46 notdan Well the directory structure that you had
16:46 notdan oh also ~/.darcs/darcsden_repos
16:47 Heffalump the darcsden-cli  repo is in ~/darcsden/darcsden-cli
16:47 Heffalump the _repos file just contains one line with the darcsden-cli repo
16:48 notdan hm
16:49 notdan I cannot reproduce the error :[
17:26 Heffalump do you have any tests of the recorded stuff?
17:26 Heffalump if so and I can run them then we can see if I have any problems
17:50 sm hi all
17:51 sm notdan: http://hub.darcs.net/co-dan/darcsden-local-take2 is your latest for me to look at, right ?
17:52 sm btw if it's easier for you to skip darcs hub I can pull from your instance
18:02 Riastradh joined #darcs
18:10 notdan Heffalump: no, I haven't done any tests for recording. I haven't settled on the interface tho
18:10 notdan oh but I guess it's a good place for unit tests
18:10 notdan sm: yes; and it's not particularly easier
18:11 notdan It's better to host it on darcshub so you can do repository comparison any time you want
18:11 sm ok cool
18:11 sm thanks for the cleanup, I'm looking at the first five patches right now
18:12 sm does "Add basic HTTP tests" depend on an unreleased hsp version ?
18:16 sm Heffalump: about your original test suite - with cabal test I see "START, binding to [http://0.0.0.0:8950/]" and then "FailedConnectionException2 "127.0.0.1" 4444 False connect: does not exist (Connection refused)" ... have I missed a step ?
18:20 sm notdan: I'm not sure you'll get that hsp patch accepted anyway, I believe it affects browser layout and would drive web designers crazy
18:20 Heffalump sm: you need to run the selenium jar (there's instructions in the README)
18:21 sm ah thanks
18:21 sm I guess this is best done on a local machine where there's a web browser
18:21 Heffalump by default it uses HTMLUnit, so you don't need a browser
18:21 Heffalump it can also use chrome and firefox
18:21 sm oh, awesome
18:22 Heffalump even though it's still a bit of a pain to run, I have put quite a lot of effort in already to removing pain :-)
18:22 sm would you happen to know if ubuntu has selenium packaged ? I'm not seeing it
18:23 alexei___ joined #darcs
18:24 Heffalump I don't know off the top of my head - I just grabbed it from their website
18:26 sm I downloaded selenium-server-standalone-2.46.0.jar .. if you've run this on gnu/linux, which java package or command did you use ?
18:26 Heffalump I can't remember which java, and just java -jar XXX.jar
18:26 sm aha, -jar does it
18:27 Heffalump looks like the Oracle jaa
18:27 Heffalump java
18:27 sm I think that's what I have installed
18:28 sm and it ran fine. I guess this is the first time I've got this far :)
18:28 sm all failing, now it looks like I need to configure couchdb somehow
18:29 sm hmm.. ERROR: Exception during CouchDB operation: src/Database/CouchDB/HTTP.hs:50:10-31: No instance nor default method for class operation pure
18:29 sm I have CouchDB-1.2.1 and ghc 7.10.1 here
18:31 Heffalump does webdriver work with GHC 7.10 now?
18:31 Heffalump it wasn't last time I tried, so I haven't tested at all
18:31 Heffalump I would expect CouchDB to just work
18:31 Heffalump but I'll have to check again
18:32 sm yes webdriver 0.6.2 is supposed to support 7.10
18:33 sm selenium always impresses me. Out of interest, what can you test via selenium+HTMLUnit ? Is that about equivalent to notdan's HTTP tests with http-client/wreq ?
18:49 sm notdan: I've pulled the first 5 + 2 fixup patches and now have 8 of 10 tests passing - yay
18:50 sm I'm using hsp 0.10.0. Are tests failing because of that ?
18:50 notdan sm: true, but I hope to either figure out a workaround or find a parser that actually works :P
18:51 notdan btw I think I broke the old selenium-based test suite :[
18:51 notdan because I meddleed with Settings.Test
18:51 sm aha
18:52 sm it sounds like I should hold off, the test patches need another round of polish ?
18:53 Heffalump you can test javascript as well
18:53 sm Heffalump: nifty
18:53 sm good to know
18:54 notdan well
18:54 notdan I was thinking about just recording a patch on top
18:54 notdan because I've been recording changes+tests together
18:54 notdan and tbh I never get selenium to work :/
18:55 notdan sm: oh yeah the wreq test are probably failing becaues of hsp
18:55 notdan i just used one patch to "fix" it http://hub.darcs.net/co-dan/hsp
18:55 sm yeah, we can't use that unless it's released
18:57 sm if you tell me how to fix the selenium tests, I could try it here
18:57 sm (if it's a small fix)
18:57 sm well let me roll back and just see
19:01 sm notdan: I still see all Heffalump's tests failing in darcsden trunk (with none of your recent changes). Or did you mean some older change ?
19:01 sm I don't think this one is your fault
19:04 notdan Yeah, they were failing for me as well. But I think I made them fail more. I switched the Settings.Test to use the local backend -- which does not support things like user creation.
19:04 notdan I guess I will make another module -- Settings.TestLocal or something -- for testing the local backend with wreq
19:05 * sm wonders how much of a priority keeping those tests running is
19:06 Heffalump well, I like them :-)
19:06 sprang left #darcs
19:06 Heffalump I'd be ok with replacing them all with http-only tests
19:06 sm ah! he heard me
19:07 Heffalump or do you mean specifically the user creation ones?
19:07 sm I like them too, just wondering about the cost/benefit for notdan
19:08 Heffalump I think it's worth having tests of both the local backend and the not-local backend
19:09 sm notdan: cabal test darcsden-test-http writes ../test/ outside of the darcsden source tree which is not so good
19:09 Heffalump I would like that fixed too :-)
19:09 notdan ah
19:09 notdan this should be fixed
19:10 Heffalump and for it not to be sensitive to it already being there
19:10 sm yes
19:10 notdan Sorry guys, I just... I don't know why I miss things like that. I guess I am just used to my working tree and I keep forgetting that what I do is not always portable
19:11 sm notdan: it's perfectly normal, that's why code review is so useful (for anyone)
19:11 notdan Heffalump: for some reason shelly just doesn't want to rm_f that directory ..
19:11 sm code review/testing
19:12 notdan OK, so I am writing code to fix those issues. Should I record it as a separate patch, or should I rebase it somewhere in the history?
19:12 alexei___ joined #darcs
19:14 sm if it's easy, I'd amend the original patches, but it's ok either way as long as I can pull them as a group
19:17 sm got to go.. I'm excited about these improvements
19:17 sm go notdan!
19:23 notdan :)
19:23 notdan One thing I am really happy about is that once we merge this blob it's going to be muuuuch easier to merge subsequent changes
19:23 notdan because the "kernel" so to say won't change much
19:34 sm Heffalump: I guess the right thing is to use selenium only for tests that need it, since the http-only tests are easier and just work
19:35 Heffalump sm: yeah, though once you've got one test with it, the rest don't cost much more
19:36 Heffalump I suspect all the pages that have a lazy load operation internally (like changes) kind of rely on the JS
19:37 notdan yeah, Javascript testing can only be done with Selenium I am afriad
19:43 notdan OK pushed the test update that fixes issues that you guys mentioned
19:43 notdan as for separating the local test settings from selenium test settings, I hope you don't mind if I record a separate patch for that
19:43 notdan I just need to figure out selenium tests myself
19:51 sm notdan: can you rebase the patches introducing the http test suite so they're together, and earliest so I can easily pull just those. And having done that much, you might want to merge a few of the fixup patches as well. At least right now, it's hard for me to tell what's going on from the history - cleaner patches will greatly assist anyone trying to compare branches, troubleshoot, darcs trackdown etc.
19:52 * sm wonders where darcs trackdown went
19:53 sm afk
19:53 sm s/merge/combine/
19:58 notdan Oh, you want me to group them together in the history?
19:58 notdan Yeah I can do that
20:00 notdan Wait a minute.. is the unique patch URL in darcsden generated from a timestamp?
20:00 notdan Because here all the URLs are the same: http://hub.darcs.net/co-dan/darcsden-local-take2/filechanges/test-http/darcsden-test-http.hs
20:01 notdan I assume that's because they've been unsuspended at the same time
20:09 alexei___ joined #darcs
20:11 notdan uh the error messages in darcs are confusing :( i propose including a CWD in every file-system related error in darcs
20:21 sm notdan: oh yes urls have resolution of 1 second, which is awkward now. I think it's in the tracker but can't find it right now
20:24 sm +1
20:25 notdan should we compute urls based on a patch's hash?
20:35 Heffalump +1
20:35 Heffalump to both the errors and the urls thing
20:42 Heffalump sm: I'm thinking of making the HsOpenSSL dependency completely conditional to make installation on Windows easier
20:42 Heffalump let me know if you have any objections
21:30 sm Heffalump: fine with me
21:33 sm notdan: yes, and ideally make the hash (or a prefix ?) be the url, so you can easily jump to it if you know the hash
23:54 lelix joined #darcs

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