00:22 <discocaml> <Ada> yeah, pin

07:52 <discocaml> <davesnx> If you want to install a specific version `opam install merlin.0.1.2.3` works also

07:53 <discocaml> <functionalprogramming> yeah but i dont want to have to check versions hehe

09:03 <discocaml> <leostera> is the behavior of the gc in hot loops documented somewhere? 🤔

09:04 <discocaml> <leostera> should a `while true do ... done` affect how often the gc gets to run?

12:31 <Square> In F# it seems you specify type arguments as "type MyType<'a,'b>(...)". Why is the "'" needed?

12:43 <companion_cube> In ocaml it's the same and it marks type variables as distinct from types

12:44 <Square> companion_cube, ok. I guess that is a motivation at least.

12:45 <companion_cube> Well yeah, the grammar isn't arbitrary :)

14:39 <discocaml> <.armael.> The compiler inserts poll points in the code so that the GC gets to run

14:40 <discocaml> <.armael.> So it shouldn’t matter whether you have an infinite loop somewhere

14:40 <companion_cube> which means every loop got slower :/

14:40 <companion_cube> oh well, that's the tradeoff of using OCaml I suppose

14:46 <discocaml> <leostera> thanks @.armael. !

16:32 <discocaml> <Ada> god i love ocaml tooling

16:36 <olle> dune too? :D

16:36 <olle> hehe

16:37 <discocaml> <Ada> dune and opam

16:37 <discocaml> <Ada> in equal measures

16:37 <discocaml> <Ada> like my two children

16:37 <discocaml> <Ada> and opam is secretly the favourite

16:46 <discocaml> <leostera> another dumb gc question 🙂 i've set up some monitoring and i can see some numbers that make sense for live/free/heap sizes in bytes. i can see my program peaks at ~531mb ram, and eventually the heap reports a size of ~254mb (70 live, 183 free) with ~44500 fragment blocks (so ~0.3mb unusable), yet somehow the resident memory of my program is at ~790mb and which is maybe accidentally close to (peak - current heap )(790-531 ~= 254 + 5). how

16:47 <discocaml> <leostera> (oh, words to bytes for a 64-bit machine)

18:12 <discocaml> <ambersaber> 怎样体验最新版本

19:35 <discocaml> <contextfreebeer> 你是用Windows嗎？OCaml5還不支持Windows吧。因為4跟5不太一樣，所以可能還沒做好Windows版本。

21:03 <discocaml> <Ada> lwt colouring is such a pain in the arse

21:03 <companion_cube> yep

21:04 <discocaml> <Kali> can't wait for effects to get good

21:16 <discocaml> <Ada> are they not already good?

21:20 <discocaml> <Kali> syntactically, i mean

21:20 <discocaml> <Kali> they are good in behavior

21:22 <discocaml> <Kali> but they are not typed either

21:23 <discocaml> <Ada> ah

21:23 <discocaml> <Ada> i've only used them through eio so far

21:23 <discocaml> <Ada> haven't got much further than hello worlds because i spend so much time working on my lwt projects haha

22:09 <discocaml> <Ada> speaking of lwt, anyboy know how i would detect that a cohttp client's closed the socket?

22:12 <discocaml> <Ada> specifically, i am trying to implement a reverse proxy type thing, and need to be able to cancel requests if the client hangs up

22:13 <discocaml> <Ada> cohttp logs this with `IO error while writing body: Unix.Unix_error(Unix.ECONNRESET, "write", "")`

22:16 <discocaml> <leostera> effects are pretty awesome already

22:21 <discocaml> <Ada> hmm, if i just attach a catch all case to cohttp's `conn_closed` callback that should work?

22:22 <discocaml> <Ada> instead of error handling just tie each response to a transfer and manually clean it up

22:22 <discocaml> <Ada> when the response has finished

22:24 <discocaml> <Ada> still some code rewrite since i'll have to switch from opium to cohttp.server directly, but thats probably better long term for maintainability

23:32 <discocaml> <regularspatula> How did you set up the monitoring to get these values ?

23:59 <discocaml> <leostera> a mix of gc.stat and htop – nothing too rigorous tbh