Potential bug: dune commands no longer work after updating macOS to Ventura

[tonyhuang12138]

Expected Behavior

After updating to the latest macOS (Version 13.3 (22E252)) today and latest dune version (3.7.0), running dune build produces the error 'Error: write(): Operation timed out' for me.

Reproduction

Below is the content of my dune file:

(executable
 (public_name platypus)
 (name toplevel)
 (libraries
  platypus
  llvm
  llvm.analysis
  llvm.executionengine
  llvm.bitwriter
  ctypes.foreign))

(rule
 (target scanner.ml)
 (deps scanner.mll)
 (action
  (chdir
   %{workspace_root}
   (run %{bin:ocamllex} -q -o %{target} %{deps}))))

(rule
 (targets parser.ml parser.mli)
 (deps parser.mly)
 (action
  (chdir
   %{workspace_root}
   (run %{bin:ocamlyacc} -v %{deps}))))

[rgrinberg]

Can you run the dune command reproducing this error with --debug-backtraces and show us the stacktrace?

[tonyhuang12138]

Hi, below is the output from running dune build --debug-backtraces, thanks!

[Alizter]

Do you have a file _build/.lock? If so, can you delete it and try again?

[tonyhuang12138]

Do you have a file _build/.lock? If so, can you delete it and try again?

Just removed the file and everything worked. Thanks!

[Alizter]

I'm going to keep this open as you shouldn't have to do this by hand. It looks like we have a case that we haven't accounted for so we will try to work out what happened here.

[Alizter]

@tonyhuang12138 just to check, are you using an m1/m2 chip?

[tonyhuang12138]

@tonyhuang12138 just to check, are you using an m1/m2 chip?

Yes, I have a M1 Pro chip.

[rgrinberg]

Are you using NFS or something? I have no idea how a write could timeout without a socket in there somewhere.

[tonyhuang12138]

I'm not entirely sure. My suspicion is that my folder is also stored on OneDrive.

[Alizter]

FTR @anmonteiro was not able to reproduce this on an m2. Unless there is more information you can give us, it seems to have been a one off. I suspect that maybe there was a permission error, but I am unable to test this.

Usually when Dune encounters a lock, it will check if there is another Dune process with ownership. If there isn't, it will go ahead and delete/overwrite the lock. I guess something caused the overwrite to wait and time out. Not sure what.

[rgrinberg]

I'm not entirely sure. My suspicion is that my folder is also stored on OneDrive.

Okay, that's probably the reason. We should just improve the error message to tell the user the lock file could not be written and we need to retry after deleting it.