[Discussion] drop 25.x support

[yyoncho]

Opening the discussion. I am now investigating whether threads could be used to improve lsp-mode responsiveness in certain cases(e. g. file watches).

[ngquerol]

I thought I'd mention that there seems to be problems with Emacs lisp threads; See the third paragraph in this article.

Now I have no idea if these problems remain in 27.x (that would still be a problem for 26.x), but if that is the case, it may be detrimental to the stability of lsp-mode. Even then, this is a low-level API and as such can deadlock Emacs instances completely.

[yyoncho]

Now I have no idea if these problems remain in 27.x (that would still be a problem for 26.x), but if that is the case, it may be detrimental to the stability of lsp-mode. Even then, this is a low-level API and as such can deadlock Emacs instances completely.

The problem(which is by design ftr) is still here - we are going to use it for background tasks and use thread-yield in the task so instead of locking emacs for 1-2 seconds emacs will be able to respond to keyboard events.

[kiennq]

If I understand correctly, Emacs only allow one thread run at a time, once you yield the thread, it will no longer proceeding, which kind of background tasks can you use for that?

[yyoncho]

@kiennq check this - yyoncho@a5facf7

[kiennq]

So you yield the thread immediately after running it so that it will be run later and can be resumed at specific points?
Why don't use generator instead of that? I think thread on Emacs is costly according to the article @ngquerol pointed out, while generator is better now.

[yyoncho]

So you yield the thread immediately after running it so that it will be run later and can be resumed at specific points?

This part might change, the code is POC.

Why don't use generator instead of that?

If it works I am fine - can you provide a prototype?

I think thread on Emacs is costly according to the article

We might use one worker thread. But more concerning is:

Update: ThreadSanitizer (TSan) quickly shows that Emacs’ threading implementation has many data races, making it completely untrustworthy. Until this is fixed, nobody should use Emacs threads for any purpose, and threads should disabled at compile time

[kiennq]

Let me try some prototype for that.

Since Emacs is basically single thread, maybe even with data races problem pointed by TSan, it's still good to use now, who knows 😄

[harismandal]

Can watchman be used for file watching? Maybe only if it's available on the system.

[yyoncho]

Can watchman be used for file watching? Maybe only if it's available on the system.

Probably, someone has to investigate that.

[kiennq]

Add one more comment to this, I surprise that this works and non-blocking

(make-thread (lambda ()
               (url-copy-file "https://speed.hetzner.de/100MB.bin" "~/tmp/a.tmp")
               (message "done!")))

Always thought that without yielding, the message will never be output.