Camelia, the Perl 6 bug

IRC log for #darcs, 2008-12-19

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

All times shown according to UTC.

Time Nick Message
00:01 codeFiend_ joined #darcs
00:12 twb joined #darcs
00:53 mornfall I guess it's a (little late) prerelease time, duh?
00:54 twb I dunno, *you're* the RM! :P
00:54 mornfall Hm, riht.
00:54 mornfall right*
00:55 mornfall I got stuck in a pub. ... *innocent smile* (but just one beer, so should be fit for doing a release nevertheless...)
01:00 mornfall 01:59:51 | xroc@ann:~/dev/darcs/branch-2.2 -> LANG=C make dist
01:00 mornfall make: *** No rule to make target `src/CommandLine.o', needed by `darcs'.  Stop.
01:00 mornfall twb: ^ btw -- doing make darcs works though...
01:00 twb mornfall: if you "make distclean && autoconf && ./configure && make dist" does it work?
01:01 mornfall twb: I did maintainer-clean, make config, ./configure, make dist...
01:01 twb Sorry, I meant maintainer-clean.
01:01 twb OK, that means the first lines of GNUmakefile are wrong.
01:01 mornfall Not urgent as far as I am concerned, just it struck me as odd. :)
01:01 twb Ah, I see.  "make dist" assumes you have already run "make"
01:02 mornfall GHC segfault. Wee.
01:02 twb I think this was already the case BEFORE I fucked with the makefile, because the semantics of dist vs. include didn't change.
01:02 twb mornfall: that's because you're using 6.8.2?
01:02 mornfall Likely
01:02 mornfall ...
01:02 twb I had to upgrade to 6.8.3 to avoid segfaults during make
01:02 twb debhttp://deb.seereason.com/debian           lenny-seereason main # haskell
01:03 twb But cabal doesn't cause the segfault :-/
01:04 PierreW_ joined #darcs
01:06 mornfall D'oh.
01:07 mornfall One day I'm gonna die of make dist.
01:07 mornfall First, it's rebuilding darcs again, second it screwed up the version.
01:09 twb Hooray for testing
01:11 mornfall twb: Hm.
01:11 mornfall twb: Why is the include autoconf.mk excluded for make dist?
01:11 mornfall That (and .depend and so on)? That's screwed up...
01:11 mornfall Especially since the version is in autoconf.mk.
01:11 mornfall No wonder it screws up the version.
01:13 mornfall Removing dist% from the filter made it produce darcs-2.2.0pre1 instead of "darcs-"... I guess that's better.
01:13 mornfall What puzzles me more is probably echo -n 2.2.0pre1 '(unknown)' > release/STATE
01:13 twb I think it's done because the "dist" rule is theoretically not supposed to build stuff.
01:15 mornfall Even better.
01:15 mornfall 02:14:53 | xroc@ann:~/dev/darcs/branch-2.2/darcs-2.2.0pre1 -> cat release/STATE
01:15 mornfall 2.1.2 (+ 378 patches)
01:15 mornfall : - (
01:16 mornfall Let's redo the maintainer clean.
01:17 twb Suits me; Debian's package can't use maintainer-clean atm because it deletes too much stuff, but distclean deletes too little
01:17 twb OK, I've found the problem.
01:17 twb "Refactor targets that prevent "include autoconf.mk" (and .depend)."
01:18 twb That goes from excluding .depend for disttest and slowdisttest, to excluding for anything starting with dist
01:20 mornfall Is it very avantgarde to make the cabal version 2.1.90?
01:21 mornfall Or maybe 2.1.8x for "pre" and 2.1.9x for "rc".
01:21 twb I don't know.
01:21 mornfall (Generally, I'd probably move to alpha/beta/rc cycle though... I wish I could say 2.2~beta2 in cabal's version: field...)
01:21 mornfall dcoutts: You don't happen to be awake, do you?
01:21 twb I might use 2.1.98.N and 2.1.99.N for alpha/pre and beta/rc respectively.
01:22 twb Since you can't express "pre" and "rc" directly in Haskell version numbers.
01:23 mornfall 2.1.(appreaches_infinity).N ;)
01:23 mornfall Weeh.
01:24 mornfall Seems to have worked.
01:26 mornfall I'd prolly love to unify the cabal/make version numbers, but it might be less work to just drop the make buildsystem altogether.
01:26 mornfall (Ie. wait for that to happen...)
01:27 mornfall Running issue709_pending_look-for-adds.sh ...    failed.
01:27 mornfall Probable reason :
01:27 mornfall . lib
01:27 mornfall issue709_pending_look-for-adds.sh: line 4: lib: Нет такого файла или каталога
01:27 mornfall Ahw.
01:27 mornfall (Sorry for the russian: no such file or directory.)
01:32 twb Your system is set to Russian?  I didn't realize you were so far east.
01:33 mornfall I'm not, really. :)
01:33 mornfall Czech Republic.
01:33 mornfall But I study Russian as a hobby.
01:33 mornfall src/Darcs/Resolution.lhs:20:13: cannot parse LANGUAGE pragma
01:33 mornfall :|
01:34 twb For some reason I thought you were in Austria.
01:34 mornfall With cabal.
01:34 mornfall Not far from Wien, actually.
01:36 mornfall I'm sort of screwed. :|
01:36 twb How so?
01:37 mornfall Well, to make good cabal sdist, I need to have a 2.1.98 tagged, and to make a good autoconf'd dist, I need to have 2.2.0pre1 tagged.
01:37 mornfall And they probably also need this to be the last tag.
01:37 mornfall *scratches head*
01:37 mornfall Well, I'll compromise.
01:38 twb You could teach cabal to translate pre into .98?
01:39 twb s/0pre/98./; s/0rc/99./
01:39 twb That would also help make it clear that .99 was a kludge and a lie
01:39 mornfall That's a little dangerous.
01:39 mornfall Also, you need to decrease the previous number by 1.
01:39 twb Oh yeah
01:40 mornfall Possibly even the rightmost non-zero number preceding rc/pre.
01:40 twb Yeah, you're right, it's fucked
01:40 twb If you're gonna break one of the two, it should be cabal, since that's still unblessed
01:41 mornfall I'm going the following route:
01:41 mornfall 1) kick the distribution into shape
01:42 mornfall 2) tag 2.1.98.1, create sdist
01:42 mornfall 3) obliterate tag, re-tag as 2.2.0pre1
01:42 mornfall 4) make dist
01:42 mornfall 5) push to public branch-2.2
01:42 twb Sounds good.
01:44 twb I have a patch prepared which should help "make dist"
01:45 mornfall Hm.
01:45 mornfall I have already recorded that "remove dist%" on my branch.
01:45 mornfall But it won't hurt to exercise the conflict resolution a little I guess.
01:48 mornfall This hspwd issue with cabal test is however unfortunate.
01:48 twb Oof, OK, I sent it
01:48 mornfall I have done sudo ln -s /bin/pwd /usr/local/bin/hspwd, but I find that a little ... curious.
01:49 twb mornfall: there's no way around that, unfortunately.
01:49 twb mornfall: if you define hspwd as a build target for cabal, it will install it
01:49 twb Cabal has no way to say "build this intermediary tool which I'll use to bootstrap later parts of the build"
01:49 twb This also affects src/preproc.hs, FWIW.
01:50 mornfall twb: Well, there is one: hspwd() { runghc hspwd.hs } might be one.
01:50 mornfall Another might be turning hspwd into a hashbang haskell script.
01:51 mornfall Don't know for preproc, but I think that's perfectly reasonable solution for hspwd.
01:51 mornfall All tests OK.
01:51 mornfall Let's see if I can get it to hackage.
01:53 twb mornfall: hmm, that stuff might work
01:53 twb mornfall: my problem is that preproc.hs actually uses large parts of src/Darcs
01:53 twb mornfall: because it needs to access the foo_help strings.
01:53 twb That doesn't affect hspwd, of course.
01:54 twb (Plan B for me, which is horrible, is to run the built darcs binary with --help to get the help output.)
01:54 mornfall twb: Many programs solve that problem by having --help-man, or --help-docbook or somesuch.
01:54 mornfall Well, many might be relative here.
01:55 twb mornfall: yeah, I know.
01:55 twb mornfall: it makes me feel bad, though
01:55 mornfall It's indeed a little dirty.
01:55 mornfall But preproc is quite dirty as well.
01:55 mornfall One way around is to separate the actual commands from their help.
01:56 twb The other problem I have is: what cabal target should build user documentation?  "cabal build" or "cabal haddock" or what?
01:56 mornfall Darcs.Help.Pull, Darcs.Help.Get, etc...
01:56 mornfall Then you can easily pull together a binary that just has the help strings in it, without pulling in the command code.
01:58 twb I think --help-man and --help-rest is the least worst approach
01:58 codeFiend joined #darcs
01:59 twb The sorting of commands into subsections is also currently done in three places: darcsman.hs, darcs.tex and "darcs help" (without an argument), and --help-man would allow me to put it in just one place.
01:59 mornfall Indeed.
02:00 mornfall Well, actually, --help and --help-rest can quite well be the same, no?
02:00 twb Unfortunately, not quite.
02:00 mornfall reST should be readable enough?
02:00 twb See issue1221
02:00 * mornfall looks
02:00 twb But apart from 1221, it pretty much just works -- certainly better than the TeX version is working now.
02:01 mornfall Hm.
02:01 twb The only bit that'd be truly ugly is the doubled colon before literal code blocks.
02:01 mornfall Can't you just indent them?
02:01 twb 1221 basically means you need "-f, --foo" instead of two columns (long and short).
02:02 mornfall Or is that just markdown?
02:02 twb mornfall: that's markdown
02:02 twb In reST, indentation is block quoting.
02:02 twb Or, I think you can prefix the whole lot with > as in Bird quoting.  But that's uglier, iMO
02:02 twb *IMO
02:03 twb Or maybe it's that the first line has to be indented and start with >>> so it looks like the Python prompt
02:03 mornfall I like bird-style for blockquotes just fine. I don't blockquote that much, I do a lot of inline code though.
02:03 twb No, I mean > for literal blocks, not blockquotes.
02:03 mornfall Well, generally just in e-mail, where it gets >'d by itself.
02:04 mornfall twb: Yeah, I'd just like more to be it the other way around. :)
02:04 mornfall But then, I'm not writing the docs.
02:04 twb I'm rather ambivalent about it.
02:04 twb But reST is in general a massive boon over markdown, IMO.
02:06 mornfall Dunno, I never had a problem with markdown. But that's just me.
02:06 twb http://docutils.sourceforge.net​/docs/dev/rst/alternatives.html
02:06 twb "A Record of reStructuredText Syntax Alternatives"
02:07 twb I have a problem with markdown having VERY limited kinds of markup, being line-oriented (no hard wrapping) and using raw HTML to do anything remotely complicated -- meaning that it's very difficult to get decent print (PDF) output, except using the evil, inferior via-HTML method (e.g. Prince XML).
02:10 mornfall twb: Trailing double space is hard wrap in markdown?
02:10 twb Huh?
02:10 mornfall You can hardwrap in markdown.
02:11 twb - this is a long
02:11 twb list
02:11 twb Markdown treats that as two separate objects IIRC.
02:11 mornfall Ah.
02:11 mornfall Hm.
02:11 wjt pandoc certainly doesn't
02:11 mornfall It apparently doesn't.
02:11 mornfall (The IkiWiki version works with lists like that. I have plenty in mine...)
02:12 markstos joined #darcs
02:12 twb OK, maybe I mis-remember.
02:12 mornfall But then, I hardly ever need anything more complicated than emphasis, links and headlines.
02:12 mornfall And bullet lists. :)
02:12 twb That is why you like markdown :-)
02:13 mornfall If I need something complicated, I usually take out TeX.
02:14 mornfall (Since HTML is vastly incapable of anything complicated anyway.)
02:14 twb TeX mainly annoys me because it has crap HTML output
02:14 mornfall True.
02:14 markstos left #darcs
02:14 mornfall It's incompatible in spirit with HTML output.
02:15 twb That, and because I need to remember to escape things like _ and ~, because they are special characters
02:15 mornfall TeX is supposed to crank out exact page layouts, and it does that well.
02:15 twb And it has poor Unicode support for CJK
02:15 twb That might be fixed in recent versions
02:16 mornfall Possible. Even Russian sort of sucks in TeX. At least in the naïve use of utf8 input.
02:17 twb With CJK you can use UTF-8, but you have to say "Hey TeX, this code you're about to see is Korean" and then later "OK, now I'm using Chinese"
02:17 twb Which I guess is partly Han unification's fault.
02:17 twb But I couldn't work out a way to say in the prelude "Hey TeX, all Han in this document is Korean"
02:18 mornfall http://repos.mornfall.net/d​arcs/darcs-2.1.98.1.tar.gz and http://repos.mornfall.net/d​arcs/darcs-2.2.0pre1.tar.gz
02:19 twb Did you grab my patch?
02:19 mornfall I'd be grateful if someone could run those two through the grinder before I announce.
02:20 mornfall twb: Nope -- but it's fine to include yours in the mainline and for pre2 I'll just pull it into branch and fix up the conflict.
02:21 twb No worries
02:21 mornfall I have essentially done the same, but just excluded dist%.
02:21 twb TBH I'm only concerned about dist use by the RM, I don't care if nobody else can get "make dist" to work
02:22 mornfall Well, I'd prefer the tarballs to be 1:1 reproducible from the repository, but that's something for the post-autoconf era I guess. :)
02:22 twb Agreed.
02:27 mornfall I'd vote to format the changelog as reST.
02:27 mornfall It'd be easier to write.
02:27 mornfall :)
02:27 mornfall (And to read.)
02:29 twb I'd actually prefer the changelog to be in either outline format (and called NEWS) or in GNU changelog format (and called ChangeLog).
02:30 twb A changelog serves two purposes: firstly, it records ALL changes; secondly, it summarizes user-visible changes.  Since "darcs changes" provides the former, I think a NEWS file would be the most useful thing (i.e. the latter).
02:38 twb One thing that annoys me about reST is the poor bibliography support
02:38 twb Oh!  And the complete lack of \appendix
02:41 twb cabal builds so much faster than make (except for the whole two-build thing now, with HSdarcs)
02:46 mornfall twb: Would you mind revising the changelog after me? I'm writing it in a little sloppy style and I'm already tired of it after 190 patches out of 280...
02:46 mornfall 380*
02:47 twb I can't look at it before Saturday, I think
02:47 twb I'm waaay behind on work-work documentation :-(
02:48 twb So I'm building those tarballs to test for you
02:49 twb make has nearly won, because libHSdarcs slows cabal down :-(
02:54 mornfall Thanks for that. I don't need the changelog done until final release, so that's still at least 3 weeks.
02:55 mornfall I'll post a draft to the list, along a short announce and plead for testing. :)
02:59 twb OK
03:15 Jerri joined #darcs
03:17 codeFiend joined #darcs
03:21 mornfall *phew*
03:22 mornfall Now we hope I haven't screwed up anything. Well, *I* for sure hope.
04:41 gwern joined #darcs
04:50 codeFiend joined #darcs
05:45 codeFiend_ joined #darcs
05:47 codeFiend__ joined #darcs
05:56 codeFiend joined #darcs
07:00 gour joined #darcs
08:43 kowey joined #darcs
08:43 kowey good morning!
08:43 * kowey eagerly installs darcs 2.2.0pre1 from the tarball
08:50 kowey ugh... http://hpaste.org/13183 <-- mornfall :-(
08:51 kowey I suspect this may be us being bitten by http://bugs.darcs.net/issue1159
10:16 lelit joined #darcs
10:40 malcolmw joined #darcs
11:28 mornfall lambdabot: @tell kowey try pulling again, please...
11:28 lambdabot Consider it noted.
11:31 C-Keen lispy: What was your suggestion for handling [DarcsFlag] again? We need to fix this in darcs-2.3. I have spotted inconsistencies with usage of the flags and we need to adress this cleanly
12:43 Igloo mornfall: When does 2.2 pre/rc testing need to be done by?
13:09 lelit` joined #darcs
13:19 gbeshers joined #darcs
13:30 mornfall Igloo: December would be good, so we can get as many necessary fixes into pre2 as possible. But generally, the earlier the better.
13:31 Igloo OK, I'm unlikely to be able to do any testing then
13:34 mornfall There's still first week of January for post-pre2 fixes, so testing that would be helpful as well, of course. But yes, the schedule is a little cramped and the holidays don't help.
16:19 gal_bolle joined #darcs
16:20 gal_bolle hi all
16:20 Igloo Yo!
18:07 kpreid_ joined #darcs
18:26 gal_bolle joined #darcs
18:26 Jerri joined #darcs
18:49 gal_bolle joined #darcs
18:51 gal_bolle joined #darcs
18:54 arjanb joined #darcs
19:45 keturn joined #darcs
19:56 Baughn joined #darcs
19:56 Baughn Is it possible to mark a patch as "local" somehow, so darcs doesn't ask about pushing it?
19:57 Baughn (I've got some local changes that I don't want to push, ever, but don't want to show up in darcs record either)
19:57 Heffalump no, but lots of people have requested this
20:07 zooko joined #darcs
20:08 zooko lelit: When I try to view a patch, trac gets an internal error because trac expects a unified diff and it is getting a darcs patch instead.
20:08 zooko http://allmydata.org/trac/tahoe/a​ttachment/ticket/534/credit.patch
20:08 codeFiend joined #darcs
20:08 zooko Hm, I guess this has nothing to do with the trac-darcs plugin, since it is just the attachment browser.
20:09 zooko Ah, I see the problem.  The attachment is named ".patch", so then trac tries to parse it as a unified diff, and if that fails, it won't show it.  Hm.
20:09 zooko Shouldn't it fall back to text/plain representation?  Or possibly ...  Um.  Hm/.
20:13 gour Baughn: put them in 'private' repo which you won't use regularly
20:46 Baughn gour: Eh?
20:46 Baughn gour: How does that work?
20:53 zooko left #darcs
20:54 gour Baughn: see http://darcs.net/manual/node6.h​tml#SECTION00671000000000000000
21:58 h0mbr3 joined #darcs
22:00 h0mbr3 left #darcs
22:57 codeFiend joined #darcs
23:17 codeFiend joined #darcs

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