Logs: liberachat/#haskell
| 2021-06-28 19:56:51 | <safinaskar> | why ghc uses llvm as one of its backends? llvm is so buggy! https://sci-hub.do/10.1145/3062341.3062343 |
| 2021-06-28 19:57:01 | <safinaskar> | and https://www.cs.utah.edu/~regehr/oopsla18.pdf |
| 2021-06-28 19:57:12 | <Rembane> | Most probably to see if it was possible. |
| 2021-06-28 19:57:27 | × | eggplantade quits (~Eggplanta@2600:1700:bef1:5e10:48c3:15b7:84fd:d26e) (Ping timeout: 272 seconds) |
| 2021-06-28 19:59:21 | <safinaskar> | this articles show that llvm authors don't care about soundness at all |
| 2021-06-28 19:59:26 | <shapr> | it also offloads a bunch of work to use LLVM |
| 2021-06-28 19:59:46 | <safinaskar> | did somebody came up with sound llvm alternative? |
| 2021-06-28 19:59:58 | <Rembane> | Yeah, which is why we've had a Cambrian explosion of programming languages the last ten years. |
| 2021-06-28 20:00:19 | <shapr> | safinaskar: ei |
| 2021-06-28 20:01:22 | <safinaskar> | Rembane: "which is why we've had a Cambrian explosion of programming languages the last ten years" - you mean this languages rely on llvm? or attempt to replace it? |
| 2021-06-28 20:01:39 | <Rembane> | safinaskar: Rely on. It seems to go quite well for them. |
| 2021-06-28 20:01:51 | <dminuoso> | I think you are mischaracterizing LLVM and the authors skillset. |
| 2021-06-28 20:01:59 | → | falafel joins (~falafel@pool-96-255-70-50.washdc.fios.verizon.net) |
| 2021-06-28 20:02:13 | → | wroathe joins (~wroathe@c-68-54-25-135.hsd1.mn.comcast.net) |
| 2021-06-28 20:02:18 | <safinaskar> | shapr: what is "ei"? |
| 2021-06-28 20:02:49 | <shapr> | Finnish, sorry |
| 2021-06-28 20:03:04 | <safinaskar> | dminuoso: this articles show the authors didn't attempt to create sound theory before creating llvm (as opposed to haskell authors) |
| 2021-06-28 20:03:10 | <shapr> | safinaskar: Yeah, there are plenty of soundness concerns with computers |
| 2021-06-28 20:03:53 | <shapr> | safinaskar: any thoughts on how to improve the whole mess? |
| 2021-06-28 20:03:56 | <Vq> | shapr: You speak finnish as well? |
| 2021-06-28 20:03:58 | <dminuoso> | safinaskar: Yes, sometimes software models have bugs. What is your point? |
| 2021-06-28 20:04:05 | <shapr> | Vq: nej tack |
| 2021-06-28 20:04:09 | Vq | is the impressed |
| 2021-06-28 20:04:11 | <dminuoso> | safinaskar: And dont make the mistaken assumption that Haskell is perfect. |
| 2021-06-28 20:04:17 | <dminuoso> | Just take a long look at the GHC bug tracker. |
| 2021-06-28 20:04:24 | <shapr> | Vq: my girfriend is taking Finnish classes and it's starting to come back to me a bit |
| 2021-06-28 20:04:26 | → | tromp joins (~textual@dhcp-077-249-230-040.chello.nl) |
| 2021-06-28 20:04:33 | <safinaskar> | shapr: "any thoughts on how to improve the whole mess" - at least implement ideas it that two papers |
| 2021-06-28 20:04:41 | <dminuoso> | GHC is so bug ridden, it's dangerous to assume that GHC generates correct code merely on the basis it compiled your code. |
| 2021-06-28 20:05:05 | <safinaskar> | shapr: write document with rigid semantics of llvm |
| 2021-06-28 20:05:24 | <safinaskar> | shapr: improve c standard (it is mess, too) |
| 2021-06-28 20:05:49 | <shapr> | Vq: I'm trying to sign up for Swedish classes, I sent the nearby school some emails in Swedish asking what level I should take. I've had no reply, sadly. |
| 2021-06-28 20:06:03 | <shapr> | safinaskar: sounds good to me |
| 2021-06-28 20:06:19 | <shapr> | also, thanks for these links, I'll read anything John Regehr has co-authored. |
| 2021-06-28 20:06:22 | <dsal> | "improve c standard" is a non-starter. |
| 2021-06-28 20:06:27 | × | dunkeln quits (~dunkeln@188.71.193.140) (Ping timeout: 268 seconds) |
| 2021-06-28 20:06:31 | <shapr> | dsal: one more won't hurt! |
| 2021-06-28 20:06:32 | <dolio> | Yeah, it's already perfect. |
| 2021-06-28 20:06:36 | <dsal> | Just turn off the machine keeping C alive. |
| 2021-06-28 20:06:52 | <dminuoso> | safinaskar: Go sign up to get on those committees and contribute then. There's no value in coming into #haskell and making wild accusations that these folks have no clue what they are doing. |
| 2021-06-28 20:06:58 | <shapr> | Is there a better bootstrapping language than C? |
| 2021-06-28 20:07:07 | <dminuoso> | safinaskar: The problems you mentioned have been brought up to the LLVM team and addressed in an accepted RFC as far as I know. |
| 2021-06-28 20:07:18 | <dsal> | shapr: C is good for what it does, just not for what people use it for. |
| 2021-06-28 20:07:23 | <safinaskar> | dminuoso: "What is your point?" - ghc should not use llvm |
| 2021-06-28 20:07:24 | <shapr> | dminuoso: oh, which RFC? |
| 2021-06-28 20:07:27 | <dminuoso> | Issues in models of all kinds arise all the time. They are not proof that the authors were incompetent, just that its very hard to miss things. |
| 2021-06-28 20:07:59 | <DigitalKiwi> | shapr: your gf is going to charm school? sus |
| 2021-06-28 20:08:04 | <dminuoso> | shapr: Sorry, *proposal |
| 2021-06-28 20:08:07 | <dsal> | safinaskar: I'm excited to see your replacement, but I suspect it's going to be a lot of work. |
| 2021-06-28 20:08:16 | <shapr> | DigitalKiwi: Nah, law school and taking Finnish classes |
| 2021-06-28 20:08:18 | <DigitalKiwi> | https://en.wikipedia.org/wiki/Finishing_school |
| 2021-06-28 20:08:20 | <DigitalKiwi> | ;p |
| 2021-06-28 20:08:24 | <shapr> | DigitalKiwi: you so funny |
| 2021-06-28 20:08:31 | shapr | hugs DigitalKiwi |
| 2021-06-28 20:08:33 | <DigitalKiwi> | ikr, once upon a time i did standup! |
| 2021-06-28 20:08:54 | <shapr> | safinaskar: I think there are other working backends for GHC, what do you think about those? |
| 2021-06-28 20:09:06 | <DigitalKiwi> | everyone laughed a lot; i'd like to think they were laughing with me |
| 2021-06-28 20:09:13 | <Vq> | Torvalds is the only Fin that got sent to charm school. |
| 2021-06-28 20:09:16 | <geekosaur> | some of which have their own bugs |
| 2021-06-28 20:09:22 | <shapr> | I'm funding https://github.com/grin-compiler/ghc-grin on patreon |
| 2021-06-28 20:09:34 | <shapr> | DigitalKiwi: as long as everyone was having fun, it's all good |
| 2021-06-28 20:09:49 | <dolio> | Vq: Wow, it worked great. |
| 2021-06-28 20:10:15 | <shapr> | safinaskar: I don't think anything is bug free, and I've rarely seen massive rewrites succeed (though it has happened!) |
| 2021-06-28 20:10:24 | <Vq> | dolio: It really did |
| 2021-06-28 20:10:28 | <safinaskar> | shapr: you will like this link, too. http://www.open-std.org/jtc1/sc22/wg14/www/docs/n1637.pdf |
| 2021-06-28 20:12:34 | <dminuoso> | safinaskar: GHC itself a lot of adhoc code that doesn't properly respect the models you imagine. There's neither a formalization of denotational semantics of Haskell, nor a proof that GHC adheres to any particular semantics. This is why GHC has a total of 18,668 reported issues, out of which 4,482 are open. |
| 2021-06-28 20:12:36 | <shapr> | safinaskar: the only approach I think might improve the soundness of a whole system would be to start with one of the bootstrapping efforts |
| 2021-06-28 20:12:48 | Rembane | waves around CakeML |
| 2021-06-28 20:12:51 | <dminuoso> | safinaskar: On the same basis of your reasoning, you shouldn't use GHC since it's an unsound bag of mess. |
| 2021-06-28 20:12:53 | × | falafel quits (~falafel@pool-96-255-70-50.washdc.fios.verizon.net) (Remote host closed the connection) |
| 2021-06-28 20:13:29 | <dsal> | There's a part of the learning curve of computers where you from "how does this work?" to "ooh, I understand now" and then look further and you're back to "how does any of this work?" Some areas are slightly better than others, but it's lucky garbage stacks all the way down. |
| 2021-06-28 20:13:48 | <dminuoso> | safinaskar: So on the scale of " |
| 2021-06-28 20:13:53 | → | falafel joins (~falafel@pool-96-255-70-50.washdc.fios.verizon.net) |
| 2021-06-28 20:14:05 | <dminuoso> | "how is llvm so buggy", ghc can compete quite well... like any other complex piece of software. |
| 2021-06-28 20:14:46 | × | _ht quits (~quassel@82-169-194-8.biz.kpn.net) (Remote host closed the connection) |
| 2021-06-28 20:15:08 | <safinaskar> | dsal: " "improve c standard" is a non-starter " - there is ongoing work. for example, here http://www.open-std.org/jtc1/sc22/wg14/www/docs/n2676.pdf folks try to give clear definition to pointer provenance, i hope they succeed |
| 2021-06-28 20:16:06 | × | hnOsmium0001 quits (uid453710@id-453710.stonehaven.irccloud.com) (Quit: Connection closed for inactivity) |
| 2021-06-28 20:17:06 | → | eggplantade joins (~Eggplanta@2600:1700:bef1:5e10:48c3:15b7:84fd:d26e) |
| 2021-06-28 20:18:14 | <safinaskar> | "I think there are other working backends for GHC, what do you think about those?" - i don't know much about them |
| 2021-06-28 20:18:15 | <dsal> | it just seems like we're really far past the point of trying to have a meaningful impact on software in general by making C a little better, but the idea seems kind of neat there. |
| 2021-06-28 20:19:29 | <dsal> | I don't want to discourage anyone from making things better. I'm just generally surprised every time I open a box at how bad the contents are. heh |
| 2021-06-28 20:23:20 | × | MQ-17J quits (~MQ-17J@2607:fb90:888a:8184:61c1:5559:6114:eee0) (Read error: Connection reset by peer) |
| 2021-06-28 20:23:37 | → | MQ-17J joins (~MQ-17J@d14-69-206-129.try.wideopenwest.com) |
| 2021-06-28 20:24:19 | <safinaskar> | shapr: "the only approach I think might improve the soundness of a whole system would be to start with one of the bootstrapping efforts" - i agree. see http://bootstrappable.org/ (they have libera chat irc), https://www.gnu.org/software/mes/ , https://github.com/oriansj/stage0 |
| 2021-06-28 20:28:43 | <shapr> | safinaskar: do you have a blog or other means of publishing your thoughts about this? |
| 2021-06-28 20:30:15 | → | aerona joins (~aerona@2600:6c54:4600:f300:1ad6:7928:ce15:e184) |
| 2021-06-28 20:31:12 | → | hnOsmium0001 joins (uid453710@id-453710.stonehaven.irccloud.com) |
| 2021-06-28 20:31:50 | × | MQ-17J quits (~MQ-17J@d14-69-206-129.try.wideopenwest.com) (Ping timeout: 265 seconds) |
| 2021-06-28 20:32:15 | → | MQ-17J joins (~MQ-17J@d14-69-206-129.try.wideopenwest.com) |
| 2021-06-28 20:35:59 | <shapr> | safinaskar: once the bootstrapping works, then maybe hardware could be more reliable. I'd like to see ECC ram in every computer. |
| 2021-06-28 20:37:03 | <safinaskar> | shapr: "do you have a blog or other means of publishing your thoughts about this?" - no. i have technical blog in russian at habr.com |
| 2021-06-28 20:37:47 | <safinaskar> | shapr: well, i like http://bootstrappable.org project, but i don't participate, i hope i will join this project in next live :) |
| 2021-06-28 20:38:59 | × | gehmehgeh quits (~user@user/gehmehgeh) (Quit: Leaving) |
| 2021-06-28 20:40:00 | <safinaskar> | shapr: you may read this: https://kukuruku.co/post/the-collapse-of-the-unix-philosophy/ . this is my post on unix, translated from russian by unknown human translator. (translation introduces small typos in shell commands) |
| 2021-06-28 20:41:13 | <zzz> | thanks for the notes, monochrom |
| 2021-06-28 20:42:16 | → | jumper149 joins (~jumper149@80.240.31.34) |
| 2021-06-28 20:45:34 | × | MQ-17J quits (~MQ-17J@d14-69-206-129.try.wideopenwest.com) (Ping timeout: 272 seconds) |
| 2021-06-28 20:45:51 | → | MQ-17J joins (~MQ-17J@d14-69-206-129.try.wideopenwest.com) |
| 2021-06-28 20:46:17 | <safinaskar> | shapr: also i have accounts on github, sourcehut and hackage (with 1 package) |
| 2021-06-28 20:47:07 | × | kayprish quits (~kayprish@46.240.143.86) (Remote host closed the connection) |
All times are in UTC.