Home freenode/#haskell: Logs Calendar

Logs: freenode/#haskell

←Prev  Next→
Page 1 .. 1111 1112 1113 1114 1115 1116 1117 1118 1119 1120 1121 .. 5022
— 502,152 events total
2020-11-07 17:31:00 <hekkaidekapus> merijn: hehe…
2020-11-07 17:31:08 <geekosaur> % :set -fprint-explicit-foralls
2020-11-07 17:31:09 <yahb> geekosaur:
2020-11-07 17:31:15 <fgaz> merijn geekosaur I fixed data-files in v2-run in my first SoC iirc
2020-11-07 17:31:17 <geekosaur> % :t (id,id)
2020-11-07 17:31:17 <yahb> geekosaur: forall {a1} {a2}. (a1 -> a1, a2 -> a2)
2020-11-07 17:31:23 <merijn> fgaz: \o/
2020-11-07 17:31:34 <merijn> fgaz: Well, now you know it wasn't for nothing!
2020-11-07 17:31:58 <fgaz> Not too much of an hack, simply an environment variable which overrides hardcoded Paths_ stuff
2020-11-07 17:31:58 <motte> hekkaidekapus: there already exists a decoder for decoding lists that contain duplicate elements - setIgnoringDuplicates
2020-11-07 17:33:13 <fgaz> merijn: shouldn't distros use the Cabal interface? Otherwise it's just layering build tools for nothing
2020-11-07 17:33:36 <fgaz> Which actually is a thing for rust and go, which don't have the Cabal vs cabal-install separation
2020-11-07 17:33:48 <hekkaidekapus> motte: Then, what’s wrong with lifting S.fromList into your list decoder? IMO, orphans (what you’re trying to do) and overlapping instances are not worth it in this case.
2020-11-07 17:34:04 <merijn> fgaz: How you build doesn't really matter
2020-11-07 17:34:42 <merijn> fgaz: The point is that the end result needs to be prefix independent, and installing executables inside "your" prefix which may rely on their own prefix independence isn't obvious
2020-11-07 17:34:46 <merijn> Actually
2020-11-07 17:34:53 <monochrom> Distros usually use Setup.hs. Setup.hs usually uses the Cabal interface.
2020-11-07 17:34:55 <merijn> How *does* that work for libraries
2020-11-07 17:35:15 <merijn> fgaz: Well...maybe the problem is already solved :p
2020-11-07 17:35:31 <monochrom> Setup.hs also has a lot of command line options that serve distro packaging needs, e.g., separate steps of building, copying.
2020-11-07 17:35:50 <dcoutts> prefix independent libs is awkward for libs with data files
2020-11-07 17:35:53 <merijn> I mean, cabal-install literally just calls into Setup.hs, so...
2020-11-07 17:36:06 <fgaz> with cabal-install you always have the store as runtime dep though
2020-11-07 17:36:06 <fgaz> it doesn't, the store is always a dependency
2020-11-07 17:36:07 <fgaz> well, unless you are lucky and you have only statically-linked no-data-files stuff
2020-11-07 17:36:09 <merijn> dcoutts: But does it even work right now?
2020-11-07 17:36:25 <dcoutts> merijn: I think it works for everything except libs with data files
2020-11-07 17:36:32 <merijn> dcoutts: ugh
2020-11-07 17:36:52 <merijn> dcoutts: Well, on the bright side, if I solve this for run-tool-depends then fixing the library case seems trivial enough
2020-11-07 17:37:08 <merijn> Since the problems seem identical
2020-11-07 17:37:37 <dcoutts> one needs a convention on how to find lib data files relative to the exe that uses them
2020-11-07 17:37:53 <matthew-> geekosaur: thanks - I didn't know about yahb
2020-11-07 17:38:18 <merijn> Well, you need a convention on how to *install* things relative to the executable, finding follows trivially from there ;)
2020-11-07 17:38:47 <merijn> dcoutts: Do you happen to know where in Cabal the Paths_ generation stuff (roughly) is?
2020-11-07 17:39:10 <fgaz> Or you could also just use the same approach as data files (Paths_), and then both can be fixed together... somehow...
2020-11-07 17:39:10 <fgaz> I think there's an issue about overhauling Paths_ generation to make it relocatable somewhere
2020-11-07 17:39:22 → Jeanne-Kamikaze joins (~Jeanne-Ka@68.235.43.102)
2020-11-07 17:39:36 <merijn> fgaz: It is already relocatable by setting the proper environment variables, I'm pretty sure?
2020-11-07 17:39:50 <fgaz> merijn: it's in Cabal/Distribution/Simple/Build/PathsModule.hs
2020-11-07 17:39:52 <merijn> Which means packagers just have to indirect via a shell script which sets those correctly
2020-11-07 17:40:10 <motte> hekkaidekapus: `S.fromList <$> list` doesn't type check though? not sure how i would use S.fromList with the decoder
2020-11-07 17:40:18 <fgaz> merijn: you'd have to set env variables for every dep with data files
2020-11-07 17:40:23 <merijn> motte: "fmap S.fromList . list" works
2020-11-07 17:40:49 <fgaz> merijn: rather, distros usually just use the Setup.hs interface to install stuff in the right places
2020-11-07 17:40:57 <motte> merijn: oh, thanks
2020-11-07 17:41:06 <monochrom> May I suggest Windows Registry? >:)
2020-11-07 17:41:22 → jamm_ joins (~jamm@unaffiliated/jamm)
2020-11-07 17:41:48 <dcoutts> merijn, fgaz: what I wanted to do (at some point in the past) was to add a flag --prefix-independent which would "do the right thing" and conflict with any othe config flags that would cause a prefix-dependence. It might require certain conventions (to handle data files etc).
2020-11-07 17:41:54 <merijn> fgaz: If that's true (I'm not sure) That's an artifact of the current (mistaken) assumption that there's only a package's install files", rather than "those of the transitive dependencies"
2020-11-07 17:42:03 × lawr3nce quits (~lawr3nce@gateway/tor-sasl/lawr3nce) (Ping timeout: 240 seconds)
2020-11-07 17:42:30 hackage yesod-auth-hashdb 1.7.1.5 - Authentication plugin for Yesod. https://hackage.haskell.org/package/yesod-auth-hashdb-1.7.1.5 (paulrouse)
2020-11-07 17:42:46 <hekkaidekapus> > S.fromList <$> pure [1, 1, 2, 3, 3] -- motte
2020-11-07 17:42:48 <lambdabot> error:
2020-11-07 17:42:48 <lambdabot> • Ambiguous type variable ‘f0’ arising from a use of ‘show_M442697496552...
2020-11-07 17:42:48 <lambdabot> prevents the constraint ‘(Show
2020-11-07 17:42:57 <merijn> dcoutts: That's seems reasonable
2020-11-07 17:43:13 <motte> hekkaidekapus: but i still need to write an overlapping instance?
2020-11-07 17:43:20 <merijn> dcoutts: I dread the thought of adding a flag to cabal(-install), though >.>
2020-11-07 17:43:49 <dcoutts> Heh, It's not that bad. This would be primarily a Cabal Setup.hs flag.
2020-11-07 17:43:52 <merijn> I don't think any mere mortal understands the weird ass indirection through Setup.hs via Cabal
2020-11-07 17:43:57 <monochrom> I have a cunning plan! The lack of --prefix=foo means prefix independence.
2020-11-07 17:44:07 <fgaz> merijn: "assumption that there's only a package's install files" isn't that true though? When using the Setup.hs interface of a specific package
2020-11-07 17:44:18 <monochrom> How do you like my idea of subtracting a flag? :)
2020-11-07 17:44:31 <merijn> monochrom: That won't work, because --prefix would install everything relative to that, which many distros don't do/want
2020-11-07 17:44:44 <dcoutts> monochrom: it needs a bit more than that. All the other file flags have to be '$prefix/' for starters.
2020-11-07 17:45:06 <merijn> fgaz: It's not, because if you install an executable in a prefix independent way the transitive dependencies linked in need to be able to find their stuff too
2020-11-07 17:45:31 × Tario quits (~Tario@201.192.165.173) (Read error: Connection reset by peer)
2020-11-07 17:45:50 <fgaz> merijn: ah yes, I thought you were referring to the current prefix-dependent way
2020-11-07 17:45:50 <dcoutts> monochrom: the benefit of an explicit flag is that it's a clear declaration of intent, so it means it's reasonable for it to impose restrictions, whereas just not specifying --prefix is a bit weak
2020-11-07 17:46:12 <fgaz> whoosh
2020-11-07 17:46:17 <fgaz> (I think)
2020-11-07 17:46:24 → DataComputist joins (~lumeng@static-50-43-26-251.bvtn.or.frontiernet.net)
2020-11-07 17:46:43 <merijn> monochrom: I like your idea of subtracting a flag IFF you field all user questions/complaints pertaining to it on all social media :)
2020-11-07 17:46:48 <merijn> Sounds like a fair deal!
2020-11-07 17:47:07 <monochrom> Well yeah once again backward compatibility strikes.
2020-11-07 17:47:40 <hekkaidekapus> motte: No because of this: list :: Decoder a -> Decoder [a]; S.fromList :: Ord a => [a] -> S.Set a; What do you think is the type of `fmap S.fromList . list`?
2020-11-07 17:47:50 <merijn> All problems with Cabal/cabal-install boil down to two categories: 1) backwards compat and 2) people hate robust future-proof design :)
2020-11-07 17:48:13 <merijn> (see the relentless moaning about listing modules and upper bounds)
2020-11-07 17:48:23 × Ariakenom quits (~Ariakenom@h-82-196-111-76.NA.cust.bahnhof.se) (Read error: Connection reset by peer)
2020-11-07 17:48:49 <geekosaur> "I never make mistakes!"
2020-11-07 17:48:53 <dcoutts> monochrom: it's not just backwards compat: asking for prefix independence really adds constraints compared to prefix-dependence.
2020-11-07 17:49:03 <dcoutts> It's not a bad thing to be explicit about that.
2020-11-07 17:51:04 → Ariakenom joins (~Ariakenom@h-82-196-111-76.NA.cust.bahnhof.se)
2020-11-07 17:53:36 × DataComputist quits (~lumeng@static-50-43-26-251.bvtn.or.frontiernet.net) (Quit: Leaving...)
2020-11-07 17:53:57 <fgaz> by the way is there any other language / build tool / package manager that can offer prefix independence and supports data files?
2020-11-07 17:53:58 <fgaz> and that does that well
2020-11-07 17:54:43 <fgaz> ...AppImage I guess :-P
2020-11-07 17:56:39 × alp quits (~alp@88.126.45.36) (Ping timeout: 260 seconds)
2020-11-07 17:57:09 <geekosaur> docker, sort of?
2020-11-07 17:57:24 <geekosaur> more realistically speaking, it's a Hard Problem
2020-11-07 17:58:56 → solonarv joins (~solonarv@astrasbourg-653-1-263-62.w92-161.abo.wanadoo.fr)
2020-11-07 18:00:01 × Eric-K quits (~Eric-K@s91904426.blix.com) ()
2020-11-07 18:02:36 → borne joins (~fritjof@200116b864092f009eed9778381c1dff.dip.versatel-1u1.de)
2020-11-07 18:02:54 × raichoo quits (~raichoo@dslb-178-009-074-104.178.009.pools.vodafone-ip.de) (Quit: Lost terminal)
2020-11-07 18:07:51 → polyphem joins (~p0lyph3m@2a02:810d:640:776c:76d7:55f6:f85b:c889)
2020-11-07 18:10:21 × knupfer quits (~Thunderbi@200116b82cc0b900204cae1ce61735e2.dip.versatel-1u1.de) (Remote host closed the connection)
2020-11-07 18:10:30 → knupfer joins (~Thunderbi@200116b82cc0b9002da38997f72b7245.dip.versatel-1u1.de)
2020-11-07 18:12:35 <Cheery> merijn: you or smb. told me about this regex library.
2020-11-07 18:13:09 <Cheery> I didk check inside it, I find the exact same implementation as mine.
2020-11-07 18:13:28 <Cheery> or oh.. it's a bit different
←Prev  Next→
Page 1 .. 1111 1112 1113 1114 1115 1116 1117 1118 1119 1120 1121 .. 5022

All times are in UTC.