julia_setup() fails

[Raoul-Kima]

I'm in the process of moving stuff to a new computer, but can't get JuliaCall to work there. So this issues came from a clean slate (not so clean anymore) where i had just installed the newest versions of Julia (via JuliaUp), R, RStudio, Renv, and then among many other R-Packages also JuliaCall.
I get issues from within a renv controlled R project (using JuliaCall version 0.17.6) but also when just opening a plain r session, just installing and loading JuliaCall and running julia_setup(). Specifying force=TRUE and rebuild=TRUE doesn't help.
First I tried with the original Julia install by JuliaUp, then I let JuliaCall install Julia itself, and when that failed too I installed Julia in a manually specified location without spaces in the path. JuliaCall now always uses that last install (and i dont know how to get rid of them as i cannot find an uninstaller and windows only knows of one, but thats a problem for another day).
I get this error message:

But I don't always get the warnings.

I tried running this in a fresh julia repl (not launched via JuliaCall, but separately) (and first accidentally used the juliaup install for that) where it complained about Suppressor and RCall missing. So i installed those. There was another eror after that which i didn't record, but is probably the same as im reporting a few lines down. I then realized that my julia repl is using a differen julia install then JuliaCall does, so I tried the same from a REPL of the newest julia install. There might have been some other tries like changing julia environment and trying to install suppressor and RCall via JuliaCall (which I assume didnt have any effect because julia_setup doesnt work). But in the end the spot where i ended up is that if i execute the command which julia_setup complains about in the separate julia REPL then i get this error:

At some point I think it also complained about something from JuliaCall missing when i ran that line in the Julia REPL. It stopped complaining about that when I installed JuliaCall in the plain (non-renv controlled) R environment.
Sometimes through all this I might also have gotten other error messages. It seemed quite inconsistent.

Overall I feel like at this point I'm just aimlessly trying out things and am lost and directionless. I need a more principled approach to find a solution. I hope someone here has an idea how to approach this problem.

[Raoul-Kima]

May be relevant: the R package library is in a path with spaces too. I tried specifying another library location and run JuliaCall from there, but that didnt help. But I don't remember the details anymore, maybe I made another mistake or one of the other things I changed later was blocking it at that time. I'm refraining from trying that again before someone answers here, because i'm not sure if that is the right direction to go, and to avoid messing things up even further.

[Raoul-Kima]

It tried revisiting this problem today. I found out that for some unknown reason it now works correctly when i manually point it to the julia install that i first made with juliaUp. This didnt originally work, which is why i created the other installations. Unfortunately it now has to be pointed manually there, but thats workable.

I can close this as solved, close it as unsolved, or leave it open for further investigation.

The other julia install (which it now uses by default) still shows the same errors as above. I tracked that down a bit (before noticing that the other install now works). So in case anyone is interested, here are the results:

In the message above one can see that its when the setup script gets included where the error occurs.
Stepping through the setup script line by line i get similar (but not identical) errors in the following places:

when including RmdStd.jl from the following line inside it:

If i ignore this and go further along through the script then i get a similar warning again:
when including JuliaObject.jl from the following line inside it:

If i again ignore this the rest of the script doesn't generate any errors.

[ViralBShah]

Can you try with a recent Julia version (and the master of this repo) and reopen if still an issue?

[Raoul-Kima]

I'm trying, but I'm not 100% that I'm following the correct procedure, can you confirm that the following steps do what you asked me to do?

1. Make a new Rstudio Project
2. Set it to be renv-managed (so i dont mess up my other stuff by trying this)
3. run this code:

When I do this ?JuliaCall tells me that I have version 0.17.6, and the last line of the code shown above results in this console output:

If you can confirm that this was the right procedure then I can open a new issue about that.

[ViralBShah]

Does setting R_LD_LIBRARY_PATH as discussed in the README help?

Also what if you just use plain R?

[Raoul-Kima]

I'm on windows, so I assume you mean the following procedure:

I've done this in Rstudio, and in pure R. With pure R I didn't use renv, and made sure to reinstall JuliaCall from github as before. In pure R I also did a second test run where I added it to the user PATH system variable via windows settings instead of modifying PATH within R (this time adding it to the end of PATH instead of the beginning). I verified that the path was correctly showing as added to the PATH environment variable within R in all cases.

In R-Studio the result was the same as before.
In pure R I got a slightly different Error (same for both runs):

Unfortunately it's cut off slightly to the right. It ends in "had status 1".

(my R-Version is 4.5.1, which was current until 2 days ago)

[Raoul-Kima]

I didn't use "installJulia=TRUE" in pure R, because it found the install created by RStudio. I can test doing so if desired. To do so I would have to know how to cleanly remove this installation. It doesn't seem to contain an uninstaller, and isn't listed under Apps in Windows as far as I can tell (there is one entry, which I assume is the install previously made by JuliaUp which I'd like to keep, since its the only one that works with JuliaCall at the moment).

[Raoul-Kima]

Something else I don't understand is what happens when I point JuliaCall to the other Julia installation made by JuliaUp by doing:

I have two RStudio projects, both renv controlled, where I tried this. In one of them it repeatably works correctly, in the other one it repeatably fails. In Both I am executing the exact same code. Both have JuliaCall 0.17.6 (can there still be version differences?). Other R-Packages may differ in version.

The error I get in the one that fails is the same as above, except that it isn't accompanied by a warning message (and the project path and julia path are different).

[Raoul-Kima]

Aparrently the reason why those two projects behave differently with the same Julia install is that (in these projects) before calling julia_setup() i do:
"Sys.setenv(JULIA_PROJECT=paste0(getwd(),"/juliaEnvironment"))"

The two projects have their own juliaEnvironment folder. The .toml files stored in those locations differ between the two projects. Overwriting this folder in the project where julia_setup() fails with the one from the other project makes it work correctly.

I probably modified these tomls in the correctly working project as part of my previous debugging attempts. Maybe they contain a clue about why a plain JuliaCall install doesn't work.