08:06 <discocaml> <Chimrod> Hello, I’m using Ocaml with cygwin/mingw with fdopen repo, and I have an error when installing the library ppx_inline_test. Can I report the error in the error directly in the janestreet bug tracker or should I check something in my config before ? (I know that the fdopen repo is deprecated now…)

08:07 <discocaml> <Chimrod> Hello, I’m using Ocaml with cygwin/mingw with fdopen repo, and I have an error when installing the library ppx_inline_test. Can I report the error directly in the janestreet bug tracker or should I check something in my config before ? (I know that the fdopen repo is deprecated now…)

08:29 <adrien> hard to tell, it's going to depend on the error

08:35 <discocaml> <Chimrod> This an unbound module :

08:35 <discocaml> <Chimrod> ```# File "src/ppx_inline_test.ml", line 85, characters 18-44:

08:35 <discocaml> <Chimrod> # 85 | let filename = File_path.get_default_path loc in

08:35 <discocaml> <Chimrod> # ^^^^^^^^^^^^^^^^^^^^^^^^^^

08:35 <discocaml> <Chimrod> # Error: Unbound module File_path

08:36 <discocaml> <Chimrod> # Hint: Did you mean Code_path?

08:36 <discocaml> <Chimrod> ```

09:40 <discocaml> <Chimrod> The line, seems to match the code in the git repo (for the last tagged version), so this looks like a dependancy error, or an update comming from another library, but I can’t figure which one

09:40 <discocaml> <Chimrod> The line seems to match the code in the git repo (for the last tagged version), so this looks like a dependancy error, or an update comming from another library, but I can’t figure which one

12:00 <discocaml> <darrenldl> in angstrom's definition of `many`, it's essentially using explicit closures to move the call stack to heap right?

12:16 <discocaml> <zen> is there a simple way to build statically linked binaries with ocaml? I've seen the OcamlPro blogpost but that approach seems complex and brittle, is that still the 'state of the art'?

12:28 <discocaml> <dinosaure> Currently, for my case, I'm able to compile statically Bob (https://github.com/dinosaure/bob) with my Cosmopolitan toolchain (https://github.com/dinosaure/esperanto). The latter show a full example of how to statically compile a project. Moreover, this is what we do for a long time for MirageOS (because we want to produce a full image which will be an operating system). The real _trick_ is to fetch all dependencies and compile the project wi

12:49 <neiluj> Hey! would you find it useful to have a SageMath-like library in OCaml?

12:50 <neiluj> I know there is Owl but it isn't as complete as Sagemath

12:51 <neiluj> most of the effort would be to write bindings to the various backends sagemath is using PARI/GP, and reimplementing a lot of Python code to ocaml...

13:03 <discocaml> <zen> That's interesting and I would be fine using that if it's 'fire-and-forget', I'm looking for a solution where I don't have to worry about writing linker flags etc... I also don't mind compiling platform specific binaries, and would like to support windows ideally

13:10 <neiluj> good to hear it! :) it would have a focus on cryptography so mostly algebra (polynomials, finite fields, elliptic curves, matrices, lattices...). It could start as a bindings to a subset of PARI/GP's C API with nice functors for modularity

13:11 <neiluj> because PARI/GP scripting language is awful!

13:13 <discocaml> <dinosaure> in the case that you want a portable binary across platform, `esperanto` is a really good candidate, the binary produced is polyglot for Windows, Mac, Linux and \*BSD 🙂

13:17 <neiluj> lol

13:39 <neiluj> how does the garbage collector run in ocaml? in a separate process?

13:45 <adrien> at the moment: stop-the-world upon allocation

13:53 <octachron> Only for minor collections, major collection is concurrent.

13:54 <octachron> but anyway, it is the domain themselves that are running the garbage collector.

13:54 <octachron> (possibly on a backup thread)

13:55 <adrien> hmm, sorry, I was being dumb and thinking about pre-5 (correct me if that was different even then)

13:57 <octachron> With the runtime lock in OCaml 4, the meaning of stop-the-world is a bit unclear.

13:59 <discocaml> <KW78> Is that correct ? : in ocaml 5, each domain runs its own GC, the minor being totally independent up to a multi-domain synchronisation. Major allocations (either directly or through the minor-major copying ) is done via a multi-threaded memory allocator (hence not related to GC per se, and also does make the independent minor possible). Not sure how major objects are collected and coordinated, I would assume some additional markers in the mark ph

14:00 <octachron> Sure, the GC acquires the global lock when running and is stopping the execution of the "OCaml world" but that is part of the normal concurrency model of OCaml 4.

14:00 <octachron> KW78, no, that sounds like you have the concurrent minor collections variant in mind which was the first prototype.

14:01 <octachron> The current version of the minor collection is a stop-the-world variant where all domains collaborate on minor collections.

14:04 <octachron> https://arxiv.org/abs/2004.11663 describes the switch to the concurrent minor collection to the stop-the-world one.

14:10 <neiluj> are there tools to automatically generate bindings to C functions? Ctypes has the necessary functionality to do it manually in a simple way so there may be tools built on top of ctypes to do build the bindings automagically?

14:10 <neiluj> thanks by the way for the pointers and explanation on ocaml's GC

14:12 <discocaml> <Chimrod> Found. I’ve downgraded ppxlix into 0.25.1 and it’s fine now

14:45 <neiluj> oh there's https://github.com/fdopen/ppx_cstubs excellent! :)