Perl 6 - the future is here, just unevenly distributed

IRC log for #darcs, 2013-12-18

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

All times shown according to UTC.

Time Nick Message
02:08 mizu_no_oto joined #darcs
03:03 whaletechno joined #darcs
04:19 mal`` joined #darcs
04:33 drostie joined #darcs
05:00 preflex_ joined #darcs
05:25 mizu_no_oto joined #darcs
06:40 lelit joined #darcs
06:44 thorkilnaur joined #darcs
07:45 bishboria joined #darcs
07:55 bishboria joined #darcs
08:01 raichoo joined #darcs
08:17 raichoo joined #darcs
09:12 bishboria joined #darcs
09:12 byorgey joined #darcs
09:12 DarkFox joined #darcs
09:12 lambdabot joined #darcs
09:20 DarkFox joined #darcs
09:21 bishboria joined #darcs
09:21 byorgey joined #darcs
09:21 lambdabot joined #darcs
10:10 schlaftier joined #darcs
11:56 whaletechno joined #darcs
12:34 amgarchIn9 joined #darcs
12:56 mizu_no_oto joined #darcs
13:13 yitz joined #darcs
13:14 yitz joined #darcs
13:43 helmut joined #darcs
13:45 helmut hi. how do I view patch messages encoded in something non-ascii? I tried DARCS_DONT_ESCAPE_8BIT, but half of them still looks garbage. Is there an encoding specified for patch messages?
13:47 helmut and furthermore, how do I pass them to darcs in a way retaining meaning? When I write them in utf8 and pass them to darcs, they look like garbage.
13:48 yitz what kind of non-ascii are you using?
13:49 yitz and which ones are helped by DARCS_DONT_ESCAPE_8BIT and which aren't?
13:50 helmut my terminal has utf8 encoding
13:50 helmut I have no clue what encoding the repository should have
13:51 yitz most do nowadays, at least outside of east asia
13:51 yitz but can you give some examples of characters that caused the two versions of the problem?
13:51 helmut repository in question at http://www.cse.chalmers.se/~nad/repos/lib/
13:52 yitz can you paste here a sample commit comment?
13:53 helmut if I do DARCS_DONT_ESCAPE_8BIT darcs log | recode utf8..latin1 then there is a commit "Added three lemmas relating Any, All and negation." which displays a negation character correctly.
13:53 helmut without the recode step, the message looks garbage
13:54 yitz i don't find any unicode character called "negation"
13:54 yitz what character is it supposed to be?
13:55 yitz trying to get the repo you linked to. it's huge...
13:55 helmut U+00A0
13:55 helmut press ctrl-c to get lazy
13:55 helmut you only need some top-most patches to view the issue
13:57 yitz A0 is NON-BREAKING SPACE
13:58 helmut ah right. that was the broken encoding
14:00 helmut I believe the desired code point there would be U+00AC and it should look like "All (<U+00AC>_ P) xs is ..."
14:00 helmut at least that would make sense conceptually
14:01 yitz ok i see. does the commit just before that have an APPROXIMATELY EQUAL TO?
14:02 helmut "before" is not a good relation in the context of darcs
14:02 yitz All (¬_ P) xs is
14:03 helmut so is there any convention as to how commit messages are to be encoded?
14:03 yitz Changed the definition of _≅_.
14:03 helmut and is this repo just broken or am I failing to spot the convention used?
14:04 yitz it looks to me like the repo is not broken - the correct unicode characters appear to be there.
14:04 yitz it looks like a darcs UI issue.
14:04 helmut yes, that changes approx equal to.
14:05 yitz are the underscores on either side of it actually supposed to be there?
14:05 helmut yes, they are!
14:05 yitz ok good, so yeah, it's a darcs UI issue.
14:06 helmut so I tried to record a patch using the PROPORTION symbol U+2237. It looked ok in my editor, when I darcs log, it comes out as <U+2237>, but when I set DARCS_DONT_ESCAPE_8BIT it comes out as "7"
14:06 yitz you could easily work around it for now by piping the output through a program that replaces U+nnnn by the actual unicode character
14:07 helmut the diagnosis of a ui-issue helps a lot. thanks!
14:07 DarkFox left #darcs
14:07 helmut I installed darcs 2.8.4 (+ 1 patch) using cabal
14:07 yitz 7 is 37 in unicode. so it looks like the comment is in utf-16, and the DARCS_DONT_ESCAPE_8BIT is causing that to confuse your utf-8 terminal.
14:08 helmut using a hexdump, I verified that the output actually turns the <U+2237> into a single \x37 without a \x22 or anything else.
14:09 yitz ok so DARCS_DONT_ESCAPE_8BIT looks pretty broken for this then
14:10 helmut cabal flags used: curl, http, static, terminfo, threaded, library, executable, color, mmap, hpc, test, optimize, warn-as-error, force-char8-encoding, libiconv
14:10 helmut should I try rebuilding darcs with slightly different flags?
14:10 yitz oh, i'm using a darcs i built ages ago. same behavior.
14:13 helmut oh wait. that's not the flags used, that's just all the flags...
14:14 yitz haha same as you, 2.8.4 (+1 patch). so not so surpising that i get the same behavior. :)
14:18 helmut I have to admit, that I don't particularly like the UI. Having been used to git log --patch --color=auto (display changes + diffs in a coloured pager), darcs seems like a step back, but it might just be my ignorance to existing features.
14:18 helmut I mean you can have color, but it vanishes when you pass it to a pager. :-(
14:20 yitz i never liked git. but it's so ubiquitous these days that most darcs work seems to be on making it integrate better with git. :(
14:21 helmut they target different audiences
14:21 helmut git stores versions, that would amount to tagging every patch in darcs
14:22 helmut what darcs does cannot be done using git
14:29 helmut DONT_ESCAPE_ANYTHING=1 also does not help
14:35 helmut it also appears that darcs treats file contents differently from patch metadata
14:36 helmut what is represented as <U+2237> in the message, looks like <U+00E2><U+0088><U+00B7> in the diff
14:40 yitz that makes sense. the file is just bytes to darcs, and those bytes just happen to be utf-8. whereas the patch name comes from what you type in the UI when you record the patch.
14:41 helmut this really sounds like a genuine bug in darcs now. it fails to correctly display utf8 messages in a utf8 terminal
14:42 helmut worse, when I darcs send -o such a patch, I can see th breakage in more detail.
14:42 helmut while the actual patch has the correct encoding, the human readable header has the broken "7"
14:43 yitz i guess you could call it a bug. i think darcs is correctly decoding the utf-8 you type when you record the patch, and correctly making that the name of the patch, in decoded Unicode.
14:43 yitz then it makes no attempt to display anything beyond ascii as a patch name in the UI.
14:43 yitz however, it does correctly show the Unicode hex code point.
14:43 helmut then DONT_ESCAPE_ANYTHING is broken
14:44 yitz so it just looks like Unicode support simply has not been implemented yet for the part of the UI that displays patch names.
14:44 yitz right, the DARCS_DONT_ESCAPE_8BIT thing i don't understand at all.
14:44 helmut and DARCS_ALWAYS_COLOR should be documented. ;-)
14:45 yitz but you don't want to just "not escape" it. you need to select an encoding - presumably by looking up the locale in effect on the output device - and encode the unicode text.
14:46 yitz it looks like that functionality just hasn't been implemented yet.
14:46 helmut ok
14:46 yitz so technically it's a missing feature. although nowadays unicode support is such a fundamental feature that i can understand you calling it a bug.
14:47 helmut for most parts darcs goes out of the way in terms of unicode support, just for messages it actively gets in the way. ;-)
14:47 yitz right
14:48 yitz anyway, for patch names at least, the information is there in a usable format. though not human-readable without further processing.
14:50 yitz oh look! it works if you use --xml
14:51 yitz xml presents its own set of problems for human readability, but the Unicode characters in the patch names get encoded nicely to UTF-8 and display on the terminal.
14:52 helmut so thank you very much for the diagnosis, I will look into the source. maybe there is an easy fix or workaround
14:58 yitz yw. gl. btw, i don't see any obvious issue in the darcs bug tracker, so perhaps you should submit it.
14:59 helmut ok.
15:09 yitz looked at the most recent source code - on the surface it appears to do the right thing. if you can build HEAD and it's fixed, great! if not - i'm stumped.
15:09 yitz (I looked in Darcs.Patch.Info and Darcs.Patch.Named)
15:10 helmut bug report sent
15:11 yitz head appears to be at v2.9.8
15:12 gh_ joined #darcs
15:17 helmut hmmm. the darcs version of darcs fails to cabal build with "unrecognized option `--extra-prog-path=...'"
15:17 helmut cabal configure anyway
15:18 dcoutts helmut: try cabal clean, it'll be because it's talking to an old compiled Setup.hs
15:18 helmut I just darcs get http://darcs.net, where should the compiled Setup.hs come from?
15:19 yitz i built it. bug still exists. bah. ok, thanks for the bug report!
15:19 mami left #darcs
15:19 helmut do I usually get a reply or anything for mailing bugs@darcs.net?
15:20 yitz i just used cabal install in a sandbox, using the latest HP plus cabal 1.18.1.2 from hackage.
15:20 yitz don't know. we'll see...
15:20 helmut maybe greylisting or something
15:23 helmut hmm. I am failing to satisfy cryptohash <0.9
15:31 gh_ helmut, I'm just guessing but your encoding problem may be linked to this: http://bugs.darcs.net/patch908
15:34 helmut gh_: I would guess not, because that patch only changes record, whereas the issue only happens with changes
15:34 helmut gh_: as yitz stated, the patch is correctly recorded.
15:34 gh_ alright
15:35 yitz gh_: you can see the correct unicode character in the patch name when you do darcs changes --xml
15:35 helmut gh_: oh that discussion happened before you joined. sorry
15:41 * helmut wonders what Seal does...
15:45 gh_ helmut, http://files.codersbase.com/thesis.pdf page 51
15:46 helmut gh_: thanks!
15:48 helmut 2352
15:58 helmut the seal stuff was a red herring for me.
15:59 helmut the issue resides somewhere inside or related to Printable
16:00 helmut maybe a description (not message as I used to call it), has a Both value with one side broken?
16:04 helmut I'm sorry, but I'll have to give up for today.
16:04 helmut in any case, I hope that at leas the bug report turns out useful
16:09 xymox joined #darcs
16:15 mizu_no_oto joined #darcs
16:22 raichoo joined #darcs
18:12 bishboria joined #darcs
18:12 byorgey joined #darcs
18:12 lambdabot joined #darcs
18:36 amgarchIn9 joined #darcs
21:25 favonia joined #darcs
21:52 rdesfo joined #darcs
22:01 favonia joined #darcs
22:24 mizu_no_oto joined #darcs
22:39 raichoo joined #darcs

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