16:41 <discocaml> <lukstafi> I really need a way to debug inside the Du `inline_tests` harness, I have heisenbug failures that don't manifest in an executable.

16:41 <discocaml> <lukstafi> *Dune (specifically expect tests).

16:47 <discocaml> <anthrax269> although i have no idea what you're working on, and you're probably not even looking for it, but at my job, these sort of test failures are sorta manifested forcefully with stress testing

16:48 <discocaml> <anthrax269> we primarily use Go, and iterating over maps in go is not deterministic, so we often do stress testing on them, to ensure we don't depend on ordering without knowing

16:48 <discocaml> <anthrax269> happened only 2 times, and i was the reason why 🙂

17:02 <discocaml> <sim642> With verbose dune output you should be able to get the executable it's internally building and running I think

17:02 <discocaml> <sim642> You could then run that manually under some debugger hopefully

18:33 <discocaml> <lukstafi> Thanks, the logs are a bit confusing but thanks to the encouragement I did find the executable and it reproduces my bug.

18:35 <discocaml> <lukstafi> (The confusing part is `.sandbox`, the executable is under the usual root `_build/default/test`.)

19:07 <discocaml> <lukstafi> There is just one test (I commented out all else). FFI callback into OCaml segfaults (I think on the macOS side when preparing the call) frorm the test runner binary, but not from standalone binary (where it works as expected).

19:13 <discocaml> <lukstafi> Now I guess I need to dig into the discussions about support for lldb on macOS... (it doesn't print OCaml frames)

19:53 <discocaml> <tornato> excuse u

19:53 <discocaml> <tornato> https://cdn.discordapp.com/attachments/436568060288172044/1369039204306124891/image.png?ex=681a689d&is=6819171d&hm=a4d43270c61b78f614a32211a5c85248a14b22f590d1ea1d012812a10be98e3a&

19:58 <discocaml> <edhebi> that's definetly a take

20:21 <discocaml> <tornato> but does prolog have higher order relations 😛

20:24 <discocaml> <tornato> https://www.metalevel.at/prolog/metapredicates

20:24 <discocaml> <tornato> huh