This commits reverts part of
commit e83c78b8c7
Author: João Távora <joaotavora@gmail.com>
Date: Mon Feb 20 22:43:50 2023 +0000
Eglot: respect user's Eldoc configuration by default
In that commit, I did what many longstanding issues and users were
suggesting and removed Eglot's override of two Eldoc user
configuration varibles.
I verified that Eglot's behaviour would stay mostly unaltered but my
tests were very incomplete. In short there is no way that Eglot can
work acceptably with the default setting of
'eldoc-documentation-strategy', which is
'eldoc-documentation-default'. So it must be changed, either globally
or locally in Eglot's minor mode.
This is true for any situation where both synchronous and asynchronous
documentation sources are present. In Eglot's case there are two
asynchronous sources which have more importance than the synchronous
source. So any other strategy except the
'eldoc-documentation-default' makes sense.
* lisp/progmodes/eglot.el (eglot--managed-mode): Set
eldoc-documentation-strategy to eldoc-documentation-compose.
This change addresses the problems reported in many Elglot reports
dating back to early 2021 at least:
https://github.com/joaotavora/eglot/issues/648https://github.com/joaotavora/eglot/issues/894https://github.com/joaotavora/eglot/issues/920https://github.com/joaotavora/eglot/issues/1031https://github.com/joaotavora/eglot/issues/1171
In one form or another, the reports point out that the multiple pieces
of information about the "thing at point" made available by the LSP
server are not all being considered by the ElDoc system.
The reason for this is Eglot setting/trampling the variables
'eldoc-documentation-strategy' and 'eldoc-documentation-functions' in
its minor more entry function.
The reason it did that is historical and is partially described in the
issues above. But, evidently, it never made much sense, because so
many people want to override it, which requires setting
'eldoc-documentation-strategy' to the non-default value
'eldoc-documentation-compose'.
The problem was made worse by the fact that setting it as usual in
either the Customize menu or their init file didn't work, requiring a
fairly complex Elisp snippet. That is now solved as of this commit.
If the user does not do any setting, then Eglot works basically the
same as before (i.e. shows only one piece of information).
It is arguable that the default value for
'eldoc-documentation-strategy' should change globally to
'eldoc-documentation-compose', but that has other subtle implications
and is not part of this commit.
* lisp/progmodes/eglot.el (eglot--managed-mode): Don't set Eldoc
variables greedily.
* lisp/progmodes/c-ts-mode.el (c-ts-base-mode): Consider a
"declaration_list" a block. (Bug#61635)
* test/lisp/progmodes/c-ts-mode-resources/indent.erts (Code): Add a
test case.
We can use the fact that 'treesit-indent-1' uses 'treesit-node-on'
when on a whitespace to set the actual current node as parent. Now we
can correctly indent the 'jsx_text' nodes. We also add some more
electric-indent-chars so that auto-indenting of jsx behaves a little
more fluently.
* lisp/progmodes/js.el (js--treesit-indent-rules): Add new rules.
(js-ts-mode): Add more indent-chars.
* lisp/progmodes/typescript-ts-mode.el
(typescript-ts-mode--indent-rules): Add new rules.
* lisp/progmodes/typescript-ts-mode.el
(typescript-ts-base-mode): Add more indent-chars and layout rules.
* lisp/progmodes/c-ts-mode.el (c-ts-mode--indent-styles): Make sure we
indent to great-grand-parent if inside an #ifdef...#endif block. If
grand-parent is root node, then don't indent one step.
(c-ts-mode--preproc-offset): New helper anchor function to calculate
indent offset.
* lisp/progmodes/c-ts-mode.el (c-ts-mode-toggle-comment-style): New
command.
(c-ts-base-mode-map): Add binding.
(c-ts-mode-set-modeline): New function.
(c-ts-mode): Set modeline.
(c++-ts-mode): Set modeline.
* lisp/progmodes/rust-ts-mode.el (rust-ts-mode--fontify-scope)
(rust-ts-mode--fontify-tail): New functions.
(rust-ts-mode--font-lock-settings): Use them instead of a lot of
more complex queries (bug#61302). Thus avoid having to create
block fontification by other features using the 'default' face.
Replace the catch-all query for 'variable' with an enumeration of
possible parent nodes.
* lisp/progmodes/xref.el (xref--insert-xrefs): Use face 'shadow' for
the line number colon instead of continuing it face (bug#61340).
(cherry picked from commit d6d25a3c22)
* lisp/progmodes/xref.el (xref-clear-marker-stack):
Clear the history correctly. Changing a lexical variable has no effect.
(cherry picked from commit dfdc0f5fb7)
Previously, defaulting to the empty string put candidates without
:sortText to the top of the list. since string-lessp is safe with nil
arguments, this makes them sort to the end instead.
* lisp/progmodes/eglot.el (eglot-completion-at-point): Simplify.
* lisp/progmodes/python.el (python-shell-buffer-substring): Add
`save-excursion' to prevent the point from moving.
* test/lisp/progmodes/python-tests.el (python-tests-should-not-move):
New helper function to assert that point does not move while calling a
function.
(python-shell-buffer-substring-*): Use
`python-tests-should-not-move'. (Bug#61463)
It looks like there are few if no other cases where we get this
particular error in the AST than with insertion of '#'.
Consider:
```
int
main()
{
|
}
```
If we only add '#' the AST yields (ERROR (ERROR)), which we can
exploit to hook onto electric indent. The end result should be the
hashtag anchored to column 0.
* lisp/progmodes/c-ts-mode.el (c-ts-mode--indent-styles): New rule.
(c-ts-base-mode): Add # to electric-indent-chars.
* lisp/progmodes/rust-ts-mode.el (rust-ts-mode--font-lock-settings):
Specify import queries to avoid clashing with scoped identifiers in
the code. Highlight attributes with font-lock-preprocessor-face.
* lisp/progmodes/rust-ts-mode.el (rust-ts-mode):
New treesit font-lock feature: 'assignment' (bug#61302).
(rust-ts-mode--fontify-pattern): Remove the node lookup step.
(rust-ts-mode--font-lock-settings): Update variable declaration
queries to match the 'pattern' child node right away. Add
highlights for 'assignment.
* lisp/progmodes/bug-reference.el (bug-reference-fontify): Wrap call
to syntax-ppss in save-match-data since it can clobber our
match-data (bug#61395).
Sign, ok, there's another edge case: else if statements. Because
"else if" is usually implemented as just another if statement nested
in the else branch, this creates additional levels that indentation
needs to ignore.
I converted c-ts-common-indent-block-type-regexp +
c-ts-common-indent-bracketless-type-regexp into a new, more flexible
variable, c-ts-common-indent-type-regexp-alist, to avoid adding yet
more variables in order to recognize else and if statements.
* lisp/progmodes/c-ts-common.el:
(c-ts-common-indent-type-regexp-alist): New variable.
(c-ts-common-indent-block-type-regexp)
(c-ts-common-indent-bracketless-type-regexp): Remove variables.
(c-ts-common--node-is): New function.
(c-ts-common-statement-offset): Use the new variable, and add the
"else if" special case. Also merge the code of
c-ts-mode--fix-bracketless-indent, because now the code is much more
succinct.
(c-ts-mode--fix-bracketless-indent): Merge into
c-ts-common-statement-offset.
* lisp/progmodes/c-ts-mode.el:
(c-ts-base-mode): Setup c-ts-common-indent-type-regexp-alist.
* test/lisp/progmodes/c-ts-mode-resources/indent.erts: New test.
Turns out I shouldn't have removed the explicit rules. Anyway, now it
indents properly.
* lisp/progmodes/c-ts-mode.el (c-ts-mode--indent-styles): Add rules.
* test/lisp/progmodes/c-ts-mode-resources/indent.erts: Add tests
Fix indentation for the semicolon in
while (str_a[i++] == str_b[j++])
;
* lisp/progmodes/c-ts-mode.el (c-ts-mode--indent-styles): New rule.
* test/lisp/progmodes/c-ts-mode-resources/indent.erts: New test.
* lisp/progmodes/eglot.el (eglot-completion-at-point): Return the
correct values in :company-kind for "EnumMember" and "TypeParameter".
The convention is to use kebab case rather than plain downcasing.
Reported in https://github.com/company-mode/company-mode/issues/1370.
* lisp/progmodes/compile.el (compilation-error-regexp-alist-alist):
Tighten regexps and simplify. Translate to rx.
* etc/compilation.txt: Add examples.
In collaboration with Jostein Kjønigsen.
