New project, new pyenv venv, eglot times out to pyright-langserver

[Cerebus]

Not sure what's going on here as I get no stderr to look at, but:

• New venv with pyenv, pyright installed along with my other standard set of packages.
• pyenv local set.
• eglot-ensure set.

I can see eglot spin up pyright; it just times out talking to it, kills the server, and exits. From messages:

[jsonrpc] Server exited with status 9
Error running timer: (error "[eglot] Timed out after 30 seconds")

If I swap out the project venv for another project's venv, everything's fine. This leads me to think it's something about the venv, but damnfino what.

Looking at eglot's events buffer, there's nothing unusual in the startup:

• The server binary is the correct one for the venv.
• rootPath and rootUri are correct.

Any ideas what to poke at?

[Cerebus]

Version splitting, breaks as of pyright 1.1.291.

[joaotavora]

I hope this is sufficient, please let me know if there's other details. I

This seems enough. I'll try to look at this this weekend

[joaotavora]

In the meantime, can you try to compile the latest Emacs master (or the 29.1 pretest) or otherwise try with the jsonrpc.el library that lives there? This behaviour could be caused by a problem that I fixed recently (in that jsonrpc.el library) and which might not have made it to the ELPA repository somehow.

[ndryden]

I just tested with the latest master and I see the same behavior as above.

[joaotavora]

I've just reproduced the problem. It seems that this new version of Pyright is mixing the stderr and stdout in the same channel. If I type simply:

pyright 2> /dev/null

to get rid of stderr (i.e. redirect it to /dev/null) I should not get any output back from the program. But with this pyright, I do. Can you try to do the same with an older version of pyright?

For comparison, pylsp behaves correctly. Just issuing pylsp gets you some stderr visible output, but pylsp 2> /dev/null is silent.

stdout is reserved for JSONRPC messages, so if you put the usual debugging junk that goes on stderr there, JSONRPC communication is affected, and here it break entirely, as Emacs is waiting for a JSONRPC response and sees some goofy stuff before that response. Maybe other JSONRPC clients can deal with this, but jsonrpc.el can't and I haven't yet encountered a LSP server in the many tens of servers that have been tested that does this.

I'd suggest investigating what changed on the pyright side and perhaps reporting this to them.

[ndryden]

Thanks for digging into this!

I believe I've traced the issue to a recent change in the Python wrapper package for Pyright and have opened an issue and suggested fix there (RobertCraigie/pyright-python#149).

Edit: This should be fixed now and presumably will show up in the next release.