Perl 6 - the future is here, just unevenly distributed

IRC log for #darcs, 2016-08-30

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

All times shown according to UTC.

Time Nick Message
00:00 sm I should try a pull from darcs.net eh
00:01 sm yes, it seems to be up to date. darcs dev has stopped, is it done ? :)
00:02 sm serious question actually. I hadn't realised things are this quiet
00:10 sm hmm. A darcs network test is asking me to enter my password
00:12 sm could any darcs command do that ?
00:23 Heffalump I haven't had much time for dev lately
00:23 Heffalump sm: hmm, probably not *any* darcs command, but quite a few if they need to lazily fetch patches from a remote ssh server
00:23 Heffalump I'm not quite sure if there's anything to control that cleanly
00:24 sm Heffalump! howdy
00:25 sm oh yes.. network tests are doing ssh to `whoami`@localhost, that might do it
00:26 Heffalump maybe they implicitly assume that the user has a public key setup that can login to their own account
00:27 sm guess I'll add it to http://bugs.darcs.net/patch1378
00:27 sm or maybe not
00:32 sm http://bugs.darcs.net/issue2504
00:43 sm Heffalump: did you say you were building darcs with GHC 8 ? Do you have patches for sandi-0.4 then ?
00:44 Heffalump not that I can see in my local copy, though I vaguely recall doing something about that
00:47 * Heffalump tries freshly building to make sure
00:47 sm cf http://bugs.darcs.net/issue2505
00:48 Heffalump ok, so probably not required for GHC 8.0 per se, but something that needs fixing
00:49 sm hmm..
00:49 * sm tries harder
00:49 Heffalump it works for me with cabal-install
00:49 mizu_no_oto joined #darcs
00:50 sm which sandi version is cabal picking ?
00:51 Heffalump hangon, I take that back, need to double-check something
00:52 Heffalump is it darcs 2.12 you need working with sandi 0.4? (2.13 doesn't either, just checking if it needs to be patched in the branch too)
00:53 sm good question, I'm not quite sure. I'm trying to get darcs back into stackage nightly. I think that means it must be a hackage release, not necessarily 2.12
00:54 Heffalump I don't think we'd want to push 2.13 to hackage
00:54 Heffalump and I think in general we aim to have the latest release working with latest software (dependencies, ghcs)
00:54 sm then perhaps a 2.12.1 ?
00:55 Heffalump yeah, just investigating what needs bumping. It looks like the last time I tried, I might still have been using --allow-newer with GHC 8 because some dependency hadn't been bumped either
00:57 Heffalump ok: so it looks like sandi 0.3.6 does work with --allow-newer, but sandi 0.4 needs real patches (as you suggest).
00:58 Heffalump sandi 0.4 was uploaded in June, so I probably hadn't noticed and had just assumed that GHC 8 would be fine when sandi 0.3.6 got an upper bound bump
01:07 sm possibly, another solution would be to get Magnus Therning to bump the upper bound on sandi-0.3.6. But that would be a stopgap
01:07 Heffalump sm: re your old question about _darcs in your home directory, is it the global cache? That normally goes in .darcs/cache though.
01:08 Heffalump the annoying thing about sandi 0.4 is that it requires base 4.7, and I think darcs 2.12 still supports base 4.6
01:09 Heffalump I remember gh noting we could bump our GHC dependency, but not sure we should do that on the 2.12 branch
01:09 sm Heffalump: I think it is, there's nothing much in .darcs. Maybe darcsden is linked with an older darcs that used _darcs
01:09 Heffalump I think that'd have to be pretty old for that to be the default, if it ever was. Can't remember though.
01:10 sm maybe it keeps using _darcs if it already exists ?
01:10 sm not sure why. I'll try renaming it
01:10 Heffalump possible
01:10 Heffalump anyway, I think nothing much will happen if you delete it/darcs can't find it, but it might start using more disk space (as it can't hardlink patches so much)
01:11 sm ha, let's try that
01:12 sm no, darcsden recreates _darcs
01:12 sm oh well
01:12 sm I shall symlink it
01:12 Heffalump hmm, weird
01:12 Heffalump I don't have a ~/_darcs on my machine that I run darcsden on
01:13 sm might be darcshub-ssh I guess
01:15 sm which is oooold
01:15 sm due to that undying ssh issue :)
01:16 sm thanks for the info Heffalump , see you
01:18 Heffalump gh_: were you thinking of dropping GHC 7.6/7.8 from the 2.12 branch, or just HEAD? It's definitely worth doing for HEAD given this sandi dependency issue (sandi <0.4 is needed for GHC 7.6, sandi 0.4 is needed for GHC 8.0, the two have incompatible APIs), but we could use CPP for the 2.12 branch
02:45 mizu_no_oto joined #darcs
03:10 Riastradh joined #darcs
06:50 castlelore joined #darcs
07:43 castlelore joined #darcs
07:49 Weltraumschaf joined #darcs
08:01 castlelore joined #darcs
08:01 castlelore joined #darcs
09:49 fr33domlover joined #darcs
10:35 f-a joined #darcs
10:36 f-a hello, I am trying to install darcs 2.12 (after upgrading to ghc 8) (debian jessie)
10:36 f-a http://paste.debian.net/plain/795056 this is what I get
11:22 fr33domlover f-a, which `cabal --version` are you using?
11:24 f-a f@x60s:~$ cabal --version
11:24 f-a cabal-install version 1.24.0.0
11:24 f-a compiled using version 1.24.0.0 of the Cabal library
11:24 f-a f@x60s:~$ ghc -V
11:24 f-a The Glorious Glasgow Haskell Compilation System, version 8.0.1
11:26 Heffalump f-a: I don't understand that error :-( lots of people are reporting it, but I can't repro it.
11:26 f-a Heffalump: mhhh, are we using same cabal/ghc?
11:27 f-a (because I started having that w/ ghc 8)
11:32 Heffalump I'm using ghc 8, yes
11:32 Heffalump cabal install 1.24.0.0
11:34 f-a Heffalump: OS?
11:35 f-a in any case, the offending line is `let Just lib = library pkgDescr` so I am going to track back form that
11:39 Heffalump f-a: Linux (NixOS)
11:39 f-a may be that debian stable is lagging behind somehow
11:39 Heffalump I think we clearly need to fix that logic so it doesn't rely on 'Just lib', but it's not obvious to me exactly how to
11:40 Heffalump and I'm confused why it's hitting some people and not others
12:03 ilbot3 joined #darcs
12:03 Topic for #darcs is now Darcs/Pijul Sprint, Helsinki, 6th-8th May http://darcs.net/Sprints/2016-05 | http://darcs.net/ | logs: http://irclog.perlgeek.de/darcs/ | darcs 2.10.3 is out http://darcs.net/Releases/2.10
12:15 mizu_no_oto joined #darcs
12:51 Big_G joined #darcs
14:02 sm morning
14:27 f-a joined #darcs
14:27 sm hello f-a
14:27 f-a ciao sm
14:27 sm yes that's odd.. it's as if the library section were missing from the cabal file sometimes
14:27 f-a very weird indeed. and I am not the only one, wonder what is it
14:28 sm is it repeatable for you ?
14:28 f-a as in, relaunch cabal install darcs?
14:28 sm yes as in it happens every time
14:29 f-a yeah
14:30 f-a and searching a bit, it seems related to ghc8 (or rather, new base)
14:30 f-a if you follow the lead, you come to Distribution.PackageDescription
14:31 f-a someone solved it by putting an allow-new=base, but it doesn't work here
14:35 sm f-a: this patch makes it dump the pkgDescr, if it helps: https://gist.github.com/anonymous/dd27223eb45342f02bf3c3cb03fcba9b
14:36 sm mine has ,library = Just ...
14:36 sm (cabal install pretty-show)
14:38 f-a sm: yeah I tried using trace and then just PutStrLn to see what it happened inside
14:39 f-a I wasn't able to understand the output xD well, let me try the patch and report back
14:41 sm osx el capitan, stack-installed ghc 8.0.1, brew-installed cabal 1.22.6.0/Cabal 1.22.4.0 here
14:42 f-a 1.24.0.0 ghc8.0.1
14:45 f-a pretty-show is quite handy, thanks for ... showing it to me, sm . http://www.ariis.it/link/t/sm.txt
14:46 sm yes it's very useful indeed (if it can parse your Show output)
14:46 sm works here with cabal 1.24.0.0 also
14:47 sm I'm using the darcs 2.12 source by the way
14:48 sm afk
14:48 sm ps is there an issue for this ?
15:01 f-a sm: yes, let me fish it
15:02 f-a http://bugs.darcs.net/issue2498 this one
15:04 f-a I wonder whether I should report it more upstream
15:05 sm where.. cabal-install ?
15:06 sm hmm. I forgot to say I am building with stack, maybe worth trying that
15:07 f-a yeah, or whatever maintains Distribution.PackageDescription & friends
15:07 sm ah! doing cabal configure, I get the problem
15:08 f-a :s
15:08 Heffalump I guess I should try the trace as well, given I don't get the problem
15:08 f-a how
15:08 f-a how is it possible sm ? I thought stack/cabal all were based on cabal-the-library
15:11 f-a sm: so with cabal it fails but not with stack?
15:11 sm yes
15:12 f-a Heffalump: do you use stack too?
15:18 sm here's the pkgDescr dump for the stack and cabal cases: https://gist.github.com/anonymous/de997950921036d309871d2c540b4486
15:20 f-a sm: can we pinpoint which cabal-the-library is stack using
15:20 sm the same one
15:20 f-a where by we I meant yo- ok thank you
15:20 sm it's the brew-installed Cabal 1.24.0.0 in both cases
15:22 f-a https://github.com/commercialhaskell/stack/issues/2192
15:23 sm is it talking about this issue ?
15:23 f-a I- well on second look probably no
15:26 sm the failing pkgDescr has a very short buildDepends list
15:27 sm and testEnabled = False, if that matters
15:33 sm f-a: does cabal configure --enable-tests work ?
15:33 f-a no
15:35 sm yes perhaps it's time for an issue at https://github.com/haskell/cabal/issues . cabal configure is clearly failing here
15:36 f-a indeed you are right, can I paste your gist? (since I am not (yet) a stack user)?
15:36 sm sure
15:37 sm oddly I see the same error with cabal configure -w ghc-7.8.4  (I don't have a 7.10 handy)
15:39 f-a hence it's cabal
15:39 f-a also I am always amazed at your knowledge of cabal
15:39 f-a every time I use it is a constand cabal <something> help | less
15:40 sm using it and troubleshooting it for a few years does that to you
15:40 f-a in the end did you accept that pull request in hledger?
15:40 sm and there's a small number of key options that you remember
15:41 sm eh which one ?
15:41 f-a the one which was discussed in the ML, entries without month indication
15:41 sm oh, no I haven't
15:42 sm there was no consensus to merge it
15:42 f-a indeed
15:44 f-a which stack version do you have, sm
15:48 sm 1.1.2
15:49 f-a thanks
15:51 sm ok I found a ghc 7.10.3, installed the deps required for configure, and cabal configure -w ghc-7.10.3 is working
15:52 sm and ghc 7.8.4 which previously was showing the error now doesn't get that far, it wants more deps
15:53 sm enough, I am done :)
15:54 f-a thanks for diagnosing sm
15:54 f-a I was going to ask where you stand in the cabal/stack download page flame, but I feel I know the answer
15:55 sm I am standing well back! out of harm's way!
15:55 f-a haha
15:56 sm I used cabal-install heavily for years, now I use stack whenever I can, I value both projects. Naturally
16:00 Heffalump f-a: I use cabal-install exclusively
16:01 f-a :s
17:30 f-a joined #darcs
18:53 byorgey joined #darcs
18:53 byorgey joined #darcs
18:55 tg joined #darcs
19:21 Riastradh joined #darcs
21:27 Heffalump f-a, sm: could you try the patch I just uploaded to issue2498? http://bugs.darcs.net/file6622/resolve-issue2498_-workaround-missing-library-specification.dpatch
21:27 Heffalump it just continues if the library is Nothing, I'd like to see what happens next
21:28 f-a sure, let me try
21:29 f-a welp, how to patch darcs without darcs xD
21:29 f-a I'll manually do it
21:33 f-a Heffalump: http://hastebin.com/iwadayosil.java
21:36 f-a maybe I should remove some folders in ~/.cabal?
21:49 Heffalump are you using a sandbox?
21:50 Heffalump I'm not sure what to make of the error. It can happen 'naturally' when you mess around with the installed package set, but it also may be that the lack of a library specification in the package description is causing further trouble, like not installing a library :-)
21:50 Heffalump if you are using a sandbox, you could blow it away and re-initialise it just to double-check
21:54 f-a no I am not Heffalump
21:55 f-a I am not using a sandbox, I mean
21:57 f-a Heffalump: I can nuke ~/.cabal/ if you want (I mean, it's a fresh installation, I don't have a bazillion packages in there)
21:58 Heffalump I think you need to nuke that and ~/.ghc
21:58 Heffalump or, you could try again using a sandbox (actually, that's another point of difference between us - I am using one)
21:58 Heffalump just running 'cabal sandbox init' first should do it
21:59 f-a just to make sure, can I ask you to provide a tarball of what you are actually testing?
22:01 Heffalump of my source tree?
22:02 f-a yes, I want to make sure we are testing the same thing (I can't darcs clone as I have no darcs ;_;)
22:02 Heffalump were you using 2.12 or HEAD?
22:03 f-a 2.12
22:03 f-a but head will do too
22:10 Heffalump ahah
22:10 Heffalump while preparing that, I think I've guessed what's wrong
22:11 Heffalump if (!flag(curl) && !flag(http))
22:11 Heffalump buildable: False
22:11 Heffalump could you delete those two lines from darcs.cabal and try again?
22:12 f-a okie, just mind that I nuked .cabal / .ghc just to make sure hence it will take some mins
22:26 Heffalump I need to go soon, but here's what I think is happening: that buildable: false condition kicks in when the solver is failing to sort out dependencies, and it ends up knocking the library out of the package description completely. If you remove it, you should instead get some error from the cabal solver.
22:29 f-a very good Heffalump . So I will do this, download latest tar. cabal unpack, remove those two lines, try
22:29 f-a thanks again for helping
22:29 Heffalump I can now reproduce the error by using cabal configure -f-curl -f-http, so I'm somewhat confident about the diagnosis. Trying to figure out the best fix now.
22:44 f-a Heffalump: still here? indeed I get an error in the solver
22:44 f-a http://paste.debian.net/796329/
22:44 Heffalump ok, cool. So I reckon --allow-newer --constraint 'sandi<0.4' will get you past that, and then it'll work
23:18 f-a checking whether to use -liconv... not needed. and then a friendly vim src/Darcs/Util/Hash.hs fixed the build, Heffalump
23:19 f-a of course the quick fix is to bump the relevant package in darcs.cabal, but I guess you meant to fix/modify/erase that if(!... too
23:32 sm omg! the evil cabal flag-flipping feature strikes again!
23:33 f-a I see no relationship of that line with my problem so even more applause.wav for the diagnosis
23:33 sm nice one Heffalump
23:34 sm why don't we see the error when stack runs cabal configure
23:34 sm maybe it's not unleashing the full power of the solver
23:36 sm our cabal allows cabal to do this, and shouldn't. But also cabal should have given a better error you'd think
23:37 sm s/our cabal/our cabal file/
23:43 f-a f@x60s:~/downloads/darcs-2.12.0$ darcs -v
23:43 f-a 2.12.0 (release)
23:43 f-a uhuhuhuhuhuh
23:43 f-a well done #darcs
23:44 sm nice f-a
23:44 sm and nice ghc bug you found elsewhere
23:45 f-a found == stumbled upon :P
23:45 f-a amazed at the speed acfoltzer & friends dissected it
23:45 sm well thank you for stubbing your toe on it so the rest of us won't have to :)

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