Closed jmercouris closed 2 months ago
@jmercouris although the PR looks good to me, it has negative consequences to Nyxt's Electron port. I need to investigate it further, but it's related to the transition to
bind-node-socket-thread
.I'll finish this PR.
Interesting. I wonder why?
I've merged the subset of the PR that didn't cause any issues for Nyxt.
I wonder why?
I'll find out but a bit later. Right now, the priority is to finish #26 and see where the Nyxt port leads us.
The current state of this PR aims to solve an important issue that we'd face in the future, but if I'd merge it right now it would make my life substantially more difficult to iterate on the Electron port.
Now is the time to finnish this PR since the goal is to fix Nyxt's prompt buffer handling (https://github.com/atlas-engineer/nyxt/tree/fix-prompt-buffer). Obviously it creates lots of sockets (due to JS calls) and so we run into EMFILE errors.
I'll work on it. Will probably add some tests to ensure that our IPC architecture is rock solid.
After a deep analysis, I conclude that the approach hinted in this PR doesn't solve issue 1 mentioned on #44.
Adds even more enhancements to cl-electron.