GNU bug report logs - #70541
track-changes-mode logs warnings (with input method, in Eglot buffer)

Previous Next

Full log

Message #50 received at 70541 <at> debbugs.gnu.org (full text, mbox):

From: Stefan Monnier <monnier <at> iro.umontreal.ca>
To: João Távora <joaotavora <at> gmail.com>
Cc: rcopley <at> gmail.com, 70541 <at> debbugs.gnu.org, Eli Zaretskii <eliz <at> gnu.org>
Subject: Re: bug#70541: track-changes-mode logs warnings (with input method,
 in Eglot buffer)
Date: Mon, 29 Apr 2024 16:27:42 -0400

> This idea is good, if I understand it correctly, though I would
> prefer if Eglot interfaced only with track-changes.el, and
> it would tell it that it should momentarily halt reports
> of changes to the server.

I guess we could add a function `track-change-inconsistent-state-p`
which Eglot could consult before calling `track-changes-fetch` and if
that returns non-nil, Eglot could reschedule the update to "later".

But that would just be a kludge usable by some packages like Eglot
because they have the luxury to reschedule.  For example, if some code
is run (e.g. via font-lock) which uses `syntax-ppss` while Quail's extra
chars are inserted, then `syntax-ppss` can't reschedule, and it may
record the resulting state in its cache after which
`inhibit-modification-hooks` may in turn prevent the cache from being
properly flushed when Quail removes its extra chars.

Removing the `inhibit-modification-hooks` would be a much
cleaner solution.  I asked Kenichi why he added that binding (back in
2005), but he hasn't replied yet.

> Can someone clarify in a simple example exactly what Eglot tells
> the LSP server as someone is inputting something with the Quail.
> I need to understand exactly where the "lie" is happening.

Here's the scenario: the user wants to insert `xᵃ` using the TeX input
method (the same problem can occur with other methods, of course), so
Emacs will receive the following inputs:

    x ^ a

but let's assume the user pauses between `^` and `a`.
At this point, the `before/after-change-functions` have been called to
announce the insertion of `x` and nothing more, but the buffer also has
an (unannounced) `^` inserted at point.

So the buffer's content as seen by `before/after-change-functions` (and
hence by the LSP server) is out of sync because of this additional `^`.

Since Quail does not call the `before/after-change-functions`, Eglot
doesn't specifically send Quail's input to the LSP server: in the above
scenario, `eglot--signal-textDocument/didChange` is called because the
`before/after-change-functions` was called in response to the previous
`x`.


        Stefan

Previous Next

GNU bug tracking system
Copyright (C) 1999 Darren O. Benham, 1997,2003 nCipher Corporation Ltd, 1994-97 Ian Jackson.