12:50 <discocaml> <lukstafi> @druimalban do you have FreeBSD? Would you mind checking if https://github.com/lukstafi/ocaml-gccjit `opam install . --deps-only --with-test; dune test` works for you? It works on my FreeBSD 14 but fails in opam-CI (FreeBSD 13.2).

16:00 <discocaml> <darrenldl> oh boy using `Unix.create_process` felt terrible

16:01 <discocaml> <druimalban> I have a FreeBSD workstation at a family member's. I can give it a shot in a week or two.

16:02 <companion_cube> @darrenldl and it's the _nice_ API

16:02 <companion_cube> the one that doesn't rely on implicit global state

16:04 <discocaml> <darrenldl> *shudders at memories of global state api*

16:07 <discocaml> <darrenldl> tho ig i only really had experience with 2: Str and api that relied on time zone variable

16:08 <companion_cube> :s :s

16:09 <companion_cube> Str really is bad

16:47 <discocaml> <xavierm02_> Hi. What's the easiest way to ensure that the type of constants and functions can be recovered from the bytecode? I could add superfluous operations whose name contains the name of the type at the typed ast stage, but I'd hope for something less ugly, and I'm not sure what to add to make the generated bytecode as easy as possible to detect (maybe external function calls?).

17:19 <Armael> hannes: what's the url of robur's git forge? I just tried to find the location of the sources for the caldav server by going to robur.coop, and I just.. couldn't- find it

17:20 <Armael> ah, git.robur.coop ; maybe worth linking in the footer of robur.coop

17:22 <companion_cube> https://github.com/ocaml/ocaml/pull/12885 awesome++

17:23 <hannes> Armael: you found it :) indeed, that website needs a refresh and some hyperlinks (to git.robur.coop and blog.robur.coop)

17:23 <Armael> ah nvm it's on github https://github.com/robur-coop/caldav

17:24 <Armael> ye the website links to https://builds.robur.coop/job/caldav for the caldav project, but on that webpage it's not obvious where the sources of the unikernel are

17:24 <Armael> (I remember wondering the same for other unikernels in builds.robur.coop )

17:27 <discocaml> <Alistair> I'm not too sure you could even encode that in the bytecode since labels are `int`s and `structured_constant`s don't keep the type name/constructor names around

17:28 <discocaml> <Alistair> I would imagine you could maybe encode it using `Lambda` before bytecomp, but I would imagine using the cmt file is the best way to recover the types for top-level functions / constants.

17:40 <discocaml> <Alistair> I'm not too sure you could even encode that in the bytecode since labels are `int`s and `structured_constant`s don't keep the type name/constructor names around (perhaps using the symtable?)

17:42 <discocaml> <xavierm02_> Using `external myfun : int -> int = "myfun"`, i get `ccall myfun` in the bytecode, so that would be detectable. It's just very annoying to add such functions everywhere they are needed, and the compiler not knowing what these functions do might disable some optimizations I would want

17:42 <discocaml> <xavierm02_> Using `external myfun : int -> int = "myfun"`, i get `ccall myfun` in the bytecode, so that would be detectable, and easily removable. It's just very annoying to add such functions everywhere they are needed, and the compiler not knowing what these functions do might disable some optimizations I would want

17:42 <discocaml> <Alistair> Oh yeah, forgot about externals. One way could be to use debug events

17:43 <discocaml> <Alistair> this contain an event module name + location

17:43 <discocaml> <Alistair> this contain an event module name + scoped location

17:45 <discocaml> <xavierm02_> Do you mean addign `-g` when compiling to get lines like `event "test.ml" 377-391`? Or is there some way to add something to the OCaml code to have messages I control?

17:45 <discocaml> <xavierm02_> Do you mean adding `-g` when compiling to get lines like `event "test.ml" 377-391`? Or is there some way to add something to the OCaml code to have messages I control?

17:46 <discocaml> <Alistair> Yeah, `-g` should give you pseudo events for each function definition

18:08 <discocaml> <xavierm02_> Ok this might work somewhat decently: to keep the type of `t`, replace it by `let fake_fun x = let _y = "mytype" in x in t`. Then locate the label that has the string, and where a closure is built from that string. Completely removing the pushed closure might be non-trivial, but I can push some chosen constant instead to allow removing the superfluous code.

18:09 <discocaml> <xavierm02_> I'd still prefer something simpler if someone has an idea 😄

18:29 <discocaml> <xavierm02_> Ok this might work somewhat decently: to keep the type of `t`, replace it by `let fake_fun x = let _y = "mytype" in x in t`. Then locate the label that has the string, and where a closure is built from that label. Completely removing the pushed closure might be non-trivial, but I can push some chosen constant instead to allow removing the superfluous code.