Using `JuliaCall` breaks for whole sesssion

[JZL]

This is an unbelievably niche issue but I thought I'd throw it out there

When I use JuliaCall to call Julia code within R, xrprof immediately starts erroring with [1]. This error continues once JuliaCall is initialized, even if running normal R code, and only works again when the session is restarted. I spent a tiny bit of time walking through the code to give a more detailed issue but I don't know ptrace that well and couldn't find anything helpful. I'm happy to run any debugging code or do some coredumps. I'm pretty sure it's worked fine in the past with python reticulate (and calling C doesn't hurt it) but I couldn't see any special handling of it within the repo so maybe Julia's FFI is more complicated

Thanks
Jonah

[1]

error: Failed to locate required R global variables in process 8026's memory. Are you sure it is an R program?
fatal: Failed to initialize R stack cursor.

[atheriel]

I don't know what could cause this, and a quick look at JuliaCall does not turn up anything suspicious. Some random thoughts:

1. Maybe Julia's JIT is involved here or there is another feature to Julia's runtime that differs from e.g. Python.
2. What does the process tree look like when you load JuliaCall versus a regular R session?

[atheriel]

Actually, now that I look at it more: JuliaCall embeds Julia using dlopen() and calls functions via symbol lookup. This is very strange to me, I'm not sure why it doesn't just link to libjulia as suggested by Julia's docs. But although strange, I don't see why this would make xrprof lose track of R's global variables.