Home liberachat/#haskell: Logs Calendar

Logs: liberachat/#haskell

←Prev  Next→
Page 1 .. 607 608 609 610 611 612 613 614 615 616 617 .. 18010
1,800,904 events total
2021-06-23 20:03:14 <davean> and it was hard to debug because it wasn't on the package - they do it in like the base package stuff
2021-06-23 20:03:24 <davean> so it was in code *refered* to, didn't show up in the nix derivation its self at all :/
2021-06-23 20:03:35 <davean> they broke it in the base infastructure
2021-06-23 20:03:36 merijn joins (~merijn@83-160-49-249.ip.xs4all.nl)
2021-06-23 20:03:42 <davean> so you really had to dig in to find it
2021-06-23 20:03:59 × derelict quits (~derelict@user/derelict) (Ping timeout: 244 seconds)
2021-06-23 20:04:24 <dminuoso> Judging from the sample I looked at, perhaps the deeper problem is the heavy reliance of "its only marked buildable if tests pass"
2021-06-23 20:04:30 <dminuoso> Which I think is an expensive threshold
2021-06-23 20:04:41 <dminuoso> Mostly, things should be buildable as its roughly aligned with stackage
2021-06-23 20:05:02 × juhp quits (~juhp@128.106.188.66) (Ping timeout: 250 seconds)
2021-06-23 20:05:20 <dminuoso> If the reality is that a good portion of hackage has broken tests, then propagating lose dependencies is a super dangerous thing because they can affect unrelated libraries
2021-06-23 20:05:26 <maerwald> Good reasons to avoid nix: bad ecosystem, low QA, too many chefs
2021-06-23 20:05:36 <davean> well, the core of *this* particular problem is they never believed you'd need an upgrade of the GHC distributed packages
2021-06-23 20:06:12 <davean> And they didn't want to consider multiple versions of those via if a nix derivation for them was used or GHC's derivation
2021-06-23 20:06:39 chaosite joins (~chaosite@user/chaosite)
2021-06-23 20:06:55 juhp joins (~juhp@128.106.188.66)
2021-06-23 20:06:55 <dminuoso> Well, multiple concurrent haskellPackage versions is horrid UX in nix at the moment.
2021-06-23 20:07:02 <dminuoso> Ideally I just want to callCabal2nix my package and be done with it
2021-06-23 20:07:06 <dminuoso> This should improve things, not worsen then
2021-06-23 20:07:20 <dminuoso> In light of all of this, I think I should really look into haskell.nix now
2021-06-23 20:07:27 <davean> yah ...
2021-06-23 20:08:14 <maerwald> haskell.nix shows that you need 10 nix engineers in a company to get anything done ;)
2021-06-23 20:09:06 <dminuoso> I dont think its that extreme, but having followed quite a bit of rabbit holes recently I am convinced you do need 2 experts at least.
2021-06-23 20:09:20 <dminuoso> Whether its worth it, you have to decidef
2021-06-23 20:09:43 <dminuoso> We are going to use nix for configuring servers now, simply because the UX and advantages are simply uncontested
2021-06-23 20:09:48 <dminuoso> But that's sort of orthogonal to the above problems
2021-06-23 20:10:03 <davean> Yah, nix is a good idea, but there are deep problems. I don't think they're with nix specificly.
2021-06-23 20:10:08 <dminuoso> Maybe, I might even think about switching back to cabal for haskell packages, and then patchelf'ing them into nix derivations
2021-06-23 20:10:13 <davean> But the ecosystem built?
2021-06-23 20:10:18 <davean> It uh, yah, has problems
2021-06-23 20:10:21 <maerwald> Yeah, easy: no. The reason IOHK went down this rabbit hole is because daedalus is installed on the users machine via nix. IMO, that's crazy.
2021-06-23 20:11:05 <dminuoso> davean: To be fair, every package distribution methods has its deep issues. I dont even want to start to discuss everything that's wrong with say rpm..
2021-06-23 20:11:18 <davean> Yah
2021-06-23 20:11:36 <maerwald> That's not a fair comparison
2021-06-23 20:11:48 <maerwald> It's not about the format or the PM
2021-06-23 20:11:50 shiraeeshi joins (~shiraeesh@109.166.57.36)
2021-06-23 20:12:01 wroathe joins (~wroathe@c-68-54-25-135.hsd1.mn.comcast.net)
2021-06-23 20:12:06 <maerwald> It's about the core developers, policies, QA
2021-06-23 20:12:10 <dminuoso> Sure
2021-06-23 20:12:28 <davean> In this case its about the philosophy of the people who are using the tool, but there are edges on the tools too, and what people do is a consiquence of the tools ergonomics too
2021-06-23 20:12:42 <davean> Its not safe to use an ecosystem you don't invest in
2021-06-23 20:12:48 <davean> sadly
2021-06-23 20:13:41 <davean> <InsertHFShillhere>
2021-06-23 20:13:46 <dminuoso> Some interesting things I recently learned, is that apparently in the go-world its super common to depend on package directly via github *tags*
2021-06-23 20:14:20 <dminuoso> When you start translating those go dependencies into nix using the common tools, that means over time your dependencies suddenly experience sha mismatches, because in the go community its normal to just move tags around too..
2021-06-23 20:14:50 <dminuoso> and apparently the same developers that came up with the go2nix (or whatevver they are called) didnt think it was a good idea to translate git refs into commit..
2021-06-23 20:14:51 <maerwald> don't remind me that I did Go professionally once... ugh
2021-06-23 20:15:11 pavonia joins (~user@user/siracusa)
2021-06-23 20:15:22 <maerwald> The only good thing about Go is that APIs are usually stable
2021-06-23 20:15:35 <dminuoso> maerwald: Correct me if Im wrong, but this all appears to stem out of this "there should be minimal effort involved in writing software or making packages"
2021-06-23 20:15:42 <dminuoso> for whatever good or bad this does
2021-06-23 20:16:02 <maerwald> It's the lack of tooling that forces people to keep their API stable
2021-06-23 20:16:04 <dminuoso> Is there some sort of hackage equivalent in go?
2021-06-23 20:16:13 <dminuoso> or Cabal-the-spec?
2021-06-23 20:16:38 <maerwald> you can import directly from github
2021-06-23 20:17:02 <maerwald> They have some vendoring tools too
2021-06-23 20:17:08 <davean> Look, we're lucky we have ok tools. They should get better - they're being made better. As a community we should value that and invest in them
2021-06-23 20:17:09 <maerwald> they're broken as well
2021-06-23 20:17:14 <dminuoso> So if github ever goes down or loses data because of stupid admins, then if your business relies on go suddenly breaks apart?
2021-06-23 20:17:24 <maerwald> yes
2021-06-23 20:17:25 <davean> just because people have a little pain with them doesn't mean they aren't good
2021-06-23 20:17:34 × MoC quits (~moc@user/moc) (Quit: Konversation terminated!)
2021-06-23 20:17:36 × jneira_ quits (~jneira_@131.red-79-155-1.dynamicip.rima-tde.net) (Ping timeout: 250 seconds)
2021-06-23 20:17:39 hnOsmium0001 joins (uid453710@id-453710.stonehaven.irccloud.com)
2021-06-23 20:17:51 <dminuoso> Oh haha, regarding github.. a really funny and sad story from a friend..
2021-06-23 20:18:08 <dminuoso> He tried to GPDR gitlab to inquire about all data they have on him.
2021-06-23 20:18:13 × norias quits (~jaredm@c-98-219-195-163.hsd1.pa.comcast.net) (Ping timeout: 268 seconds)
2021-06-23 20:18:26 <maerwald> Well, my opinion is that PVP and semver are the reasons for API anarchy
2021-06-23 20:18:28 <dminuoso> In the jungle of Gitlab this request turned from "GPDR inquiry" to "GPDR delete"
2021-06-23 20:18:45 <dminuoso> And his account and all projects, commits, messages... *everything* was irrecoverably deleted.
2021-06-23 20:19:20 <davean> Holy, why?
2021-06-23 20:19:29 <dminuoso> Ask gitlab
2021-06-23 20:19:33 <davean> well, i backup stuff
2021-06-23 20:19:41 <maerwald> They have incompetent admins
2021-06-23 20:19:46 <JSharp> that's painful. :/
2021-06-23 20:19:47 <maerwald> Not the first incident
2021-06-23 20:20:36 <dolio> Because that's the easiest way to comply?
2021-06-23 20:20:51 <dminuoso> "we dont have anything on you"
2021-06-23 20:21:04 × peterhil_ quits (~peterhil@dsl-hkibng32-54f849-252.dhcp.inet.fi) (Ping timeout: 250 seconds)
2021-06-23 20:22:41 peterhil joins (~peterhil@dsl-hkibng32-54f849-252.dhcp.inet.fi)
2021-06-23 20:24:29 × mc47 quits (~mc47@xmonad/TheMC47) (Remote host closed the connection)
2021-06-23 20:24:53 mc47 joins (~mc47@xmonad/TheMC47)
2021-06-23 20:25:04 DNH joins (~DNH@8.43.122.6)
2021-06-23 20:25:50 × mc47 quits (~mc47@xmonad/TheMC47) (Client Quit)
2021-06-23 20:25:57 <dminuoso> maerwald: At any rate, its a bad sad that go is in that state. Overall go seems like a cute language for making minimal tools. This discussion immediately reminded me of https://gnu.support/gnu-emacs/What-is-wrong-with-MELPA-on-Microsoft-Github.html
2021-06-23 20:26:17 <dminuoso> All of this fully applies to the go-ecosystem as well, it seems.
2021-06-23 20:26:30 <dolio> Also javascript, probably.
2021-06-23 20:26:40 <maerwald> Go is good for microservices, nothing more
2021-06-23 20:27:15 <dminuoso> Ive seen a bunch of CLI tools written in go as well, it seems fairly suitable for such tasks too
2021-06-23 20:27:37 <dminuoso> Most recent thing is `morph`, which is how we're going to deploy nixos machines with. :)
2021-06-23 20:27:47 <maerwald> python is better at cli tools imo
2021-06-23 20:28:40 <dminuoso> perhaps yeah, guess it depends what you have in your toolkit
2021-06-23 20:28:52 <dminuoso> I dont have any python or go experience, so my cli tools are written in Haskell
2021-06-23 20:29:21 <dminuoso> But with Haskell some common tasks do require 5 lines of code, that should really just take 1.
2021-06-23 20:30:31 <maerwald> I'm not a huge fan of optparse-applicative. Too many details to deal wits
2021-06-23 20:30:42 × geekosaur quits (~geekosaur@xmonad/geekosaur) (Killed (NickServ (GHOST command used by allbery_b)))
2021-06-23 20:30:42 allbery_b joins (~geekosaur@xmonad/geekosaur)
2021-06-23 20:30:51 allbery_b is now known as geekosaur
2021-06-23 20:31:01 <maerwald> With TH you could build something better, but then you need TH... ugh
2021-06-23 20:31:09 × _ht quits (~quassel@82-169-194-8.biz.kpn.net) (Remote host closed the connection)

All times are in UTC.