The group parameter nnselect-always-regenerate causes the list of
articles in the group to be generated each time it is needed. For this
to work reliably the list of articles has to be generated at the
appropriate time and to have a reproducible ordering.
* lisp/gnus/gnus-search.el (gnus-search-run-search): For nnselect
groups if the article list has not been stored, regenerate it.
* lisp/gnus/nnselect.el (nnselect-generate-artlist): Sort the
generated list of articles by RSV, group, and number. Store the
artlist after generation. When the new optional argument INFO is
non-nil, update the group info.
(nnselect-compress-artlist, nnselect-uncompress-artlist): Preserve the
article list ordering.
(nnselect-get-artlist): Return nil when nnselect-always-regenerate is t.
(nnselect-store-artlist): Store the group active range along with the
artlist. Don't keep the artlist if nnselect-always-regenerate is t.
(nnselect-request-group): The full article list isn't needed at this
stage, only the active range.
(nnselect-retrieve-headers): Regenerate the article list if there
is no stored value. Inhibit gnus-demon while retrieving headers.
(nnselect-request-group-scan): Don't generate the article list when
nnselect-always-regenerate is t since it will be generated again later
on.
(nnselect-request-create-group): Allow the artlist to be passed as an
argument to the function. Update the group info and store the artlist.
Allow thread referral to use search whenever possible, displaying the
results in the current summary buffer if possible and a new nnselect
buffer if not.
* lisp/gnus/nnimap.el (nnimap-request-thread): Obsolete function.
* lisp/gnus/gnus-search.el (gnus-search-thread): Allow detailed
specification of how/where to search. Add found articles to the
current summary buffer if possible, or create a new ephemeral nnselect
group if not.
* lisp/gnus/gnus-sum.el (gnus-refer-thread-use-search): Allow a list
of servers and groups to search.
(gnus-summary-refer-thread): Find thread-related articles by using a
backend-specific method, gnus-search, or retrieving nearby headers in
the current group.
* lisp/gnus/nnselect.el (nnselect-search-thread): Obsolete function.
(nnselect-request-thread): Allow thread referral from nnselect groups.
* doc/misc/gnus.texi (Finding the Parent): Document changes to thread
referral.
* src/treesit.c (safe_assq)
(treesit_traverse_get_predicate): New functions.
(treesit_traverse_validate_predicate)
(treesit_traverse_match_predicate): Support symbols.
(Ftreesit_search_subtree)
(Ftreesit_search_forward)
(Ftreesit_induce_sparse_tree)
(Ftreesit_node_match_p): Move validation down so we can pass LANGUAGE
to it.
* src/treesit.c:
(treesit_initialized): Make static.
(treesit_find_override_name): Add check for XCAR (tail).
(Ftreesit_parser_set_included_ranges): Fix comment.
(treesit_recursion_limit): Change to a compile time
constant.
(treesit_symbol_to_c_name): Precompute the length.
(Ftreesit_pattern_expand): Use predefined symbols.
(treesit_cursor_helper)
(Ftreesit_search_subtree)
(Ftreesit_induce_sparse_tree): Update treesit_recursion_limit.
(syms_of_treesit): New symbols.
treesit-search-forward and friends now accept more shapes for PRED,
make use of it in navigation functions.
* lisp/treesit.el (treesit-node-top-level): Use treesit-node-match-p.
(treesit--thing-unpack-pattern): Remove function.
(treesit-beginning-of-thing)
(treesit-end-of-thing): Remove PRED argument.
(treesit--things-around): Remove PRED argument, use
treesit-node-match-p.
(treesit--top-level-thing): Remove function.
(treesit--navigate-thing): Remove PRED argument.
(treesit-thing-at-point): Update docstring, don't unpack PATTERN.
* test/src/treesit-tests.el:
(treesit--ert-test-defun-navigation): Don't unpack pattern.
Make `defvaralias` signal an error upon attempts to create variable
alias cycles. This detects errors earlier and makes the alias
traversal during execution simpler and faster since no cycle detection
is needed elsewhere.
Now variable and function aliases are handled identically in these
respects.
* src/lisp.h (indirect_variable): Remove declaration.
* src/data.c (indirect_variable): Remove.
(Findirect_variable): Update doc string. Simplify alias resolution.
(Fboundp, find_symbol_value, set_internal, default_value)
(set_default_internal, Fmake_variable_buffer_local)
(Fmake_local_variable, Fkill_local_variable, Flocal_variable_p)
(Flocal_variable_if_set_p, Fvariable_binding_locus):
* src/buffer.c (buffer_local_value):
* src/eval.c (specbind): Simplify variable alias resolution.
(Fdefvaralias): Update doc string. Check for cycles.
* doc/lispref/variables.texi (Variable Aliases):
Mention that `defvaralias` can signal `cyclic-variable-indirection`
but `indirect-variable` cannot.
* etc/NEWS: Announce the change.
* test/src/eval-tests.el (eval-tests-defvaralias): New test.
This fixes bug #62386.
* lisp/progmodes/cc-engine.el (c-forward-over-compound-identifier): Don't
move forward over whitespace following the identifier.
(c-forward-primary-expression): Add parameter stop-at-end meaning don't move
forward over whitespace after the construct when non-nil. Don't recognise a
primary expression when an open brace follows a parenthesized expression.
(c-forward-constraint-clause): Extracted from c-forward-c++-requires-clause.
Add parameter stop-at-end as above.
(c-forward-c++-requires-clause): New stop-at-end parameter as above. Call the
new function c-forward-constraint-clause.
(c-forward-concept-fragment, c-looking-at-concept)
(c-in-requires-or-at-end-of-clause, c-c++-vsemi-p): New functions.
(c-guess-basic-syntax): New CASE 5A.7: "defun" open in a requires expression.
CASE 5F: Close of a non-class declaration level block: Move to earlier in the
function.
CASE 5D: Also check for being in or at end of a constraint.
New CASE 20: A C++ requires sub-clause.
New CASE 16G: The closing brace of a requires clause.
New CASE 17J: First "statement" inside a requires "function".
(c-forward-primary-expression, c-forward-declarator, c-forward-decl-or-cast-1)
(c-looking-at-or-maybe-in-bracelist): Amend the method of detecting end of
symbol "requires" (aka c-fun-name-substitute-key).
* lisp/progmodes/cc-fonts.el (c-get-fontification-context): Amend the method
of detecting end of symbol "requires".
* lisp/progmodes/cc-langs.el (c-at-vsemi-p-fn): Change the C++ entry to
c-c++-vsemi-p.
(c-fun-name-substitute-key): Change to an unadorned regexp.
* lisp/progmodes/cc-vars.el (c-offsets-alist): Add new syntactic symbol
constraint-cont.
* doc/misc/cc-mode.texi (Syntactic Symbols): Add an entry for contraint-cont.
(Constraint Symbols): New node under Syntactic Symbols.
* src/treesit.c (Ftreesit_node_check, Ftreesit_pattern_expand)
(Ftreesit_query_capture, treesit_traverse_validate_predicate)
(treesit_traverse_match_predicate):
Use BASE_EQ instead of EQ where this is obviously correct.
This is relevant for bug #58558, although it does not fix it. Due to a wrong
ordering of with-current-buffer and a let form, the function overwrote the
global value of parse-sexp-lookup-properties and two other variables.
* lisp/progmodes/cc-defs.el (c-emacs-features): Change the nesting of
with-current-buffer and let so that the let bindings get used.
* src/treesit.c:
(treesit_traverse_validate_predicate): Don't accept symbols.
(treesit_traverse_match_predicate): Don't accept symbols, and use
correct variable for the regexp and pred check.
* test/src/treesit-tests.el:
(treesit-search-forward-predicate): Fix the test.
* lisp/gnus/nnselect.el (nnselect-push-info): Don't update backend
marks when quit-config is not nil since gnus-update-marks has already
been called. Move checking for unread articles outside the
gnus-atomic block so it may be interrupted. Replace let* with let.
Cleanup code.
Right now we support regexp strings and predicate functions for the
PRED argument. This change adds support for (not ...) (or ...)
and (regexp . pred) predicates.
I still need to find a place to document the supported shapes of a
predicate.
* src/treesit.c (treesit_traverse_validate_predicate): New function.
(treesit_traverse_match_predicate): Support more predicate shapes.
(treesit_search_dfs):
(treesit_search_forward)
(treesit_build_sparse_tree): Fix docstring (unrelated to this change).
(Ftreesit_search_subtree)
(Ftreesit_search_forward)
(Ftreesit_induce_sparse_tree): Use the new function to validate
predicate shape.
(syms_of_treesit): New error Qtreesit_invalid_predicate.
* test/src/treesit-tests.el:
(treesit--ert-search-setup): Add edebug declaration.
(treesit-search-forward-predicate)
(treesit-search-forward-predicate-invalid-predicate): New tests.
Earlier we switched to using cursors rather than nodes to traverse the
parse tree. Because cursors need cleanup, we have to catch signals
thrown by the predicate functions and free the cursor. Failing to do
this will result in leaking the cursor whenever the predicate function
signals in a search function.
This change fixes the leak.
* src/treesit.c (treesit_traverse_cleanup_cursor): New function.
(Ftreesit_search_subtree)
(Ftreesit_search_forward)
(Ftreesit_induce_sparse_tree): Catch signals.
After 2ce27563ec, treesit--navigate-things takes a TACTIC argument
instead of using treesit-defun-tactic, so the tests need to change
from binding treesit-defun-tactic to passing the tactic argument,
which is what this change does.
* test/src/treesit-tests.el:
(treesit--ert-insert-and-parse-marker): New argument TACTIC.
(treesit-defun-navigation-nested-1)
(treesit-defun-navigation-nested-2)
(treesit-defun-navigation-nested-3)
(treesit-defun-navigation-top-level): Pass TACTIC argument.
Always have inlining of functions defined by `cl-defsubst` let-bind
arguments instead of making incorrect guesses when it might be safe to
substitute them and then botching the substitution.
This change generally results in better and safer code for all
callers, in particular `cl-defstruct` constructors, accessors and
mutators.
* lisp/emacs-lisp/cl-macs.el (cl-defsubst): Remove outdated comment.
(cl--defsubst-expand): Simplify: always let-bind.
(cl--sublis): Remove.
(cl-defstruct): Simplify: remove old hack that is no longer needed.
bug#62816
This is about a particular value for 'eldoc-documentation-strategy',
'eldoc-documentation-compose'. Its helper
'eldoc--documentation-compose-1' was buggy. It created the callback
for all the backends in 'eldoc-documentation-functions', but arranged
so that it could potentially be invoked immediately and trigger
display, half-defeating the purpose of the "patience" and causing
blinking in the echo area.
Now it creates all the callbacks beforehand and only then passes them
to the corresponding members of eldoc-documentation-functions. This
sets up the correct state in eldoc--invoke-strategy.
* lisp/emacs-lisp/eldoc.el (eldoc--documentation-compose-1):
Delete.
(eldoc-documentation-compose)
(eldoc-documentation-compose-eagerly): Rework.
This sweeping fix has been planned for a while, but a user recently
hit this bug as described in
https://github.com/joaotavora/eglot/discussions/1206.
More and more servers today are "multi-language", meaning can handle
more than one file type. This relies on the ':languageId' string
being set to the correct value for every buffer managed by Eglot
(TextDocumentItem in LSP parlance).
Previously this string was calculated based on an imperfect heuristic
and was wrong quite often. Many servers don't even care but some
others do, so we have to fix it.
* lisp/progmodes/eglot.el (eglot-lsp-server): Remove slots
'major-modes' and 'language-id'. Add slot 'languages'.
(eglot--major-modes, eglot--language-ids): New helpers.
(eglot--lookup-mode): Simplify or maybe complicate.
(eglot--guess-contact): Use new eglot--looup-mode. Change return
value.
(eglot): Rework docstring.
(eglot-reconnect): Use eglot--language-ids, not id.
(eglot--connect): Setup eglot--languages slot in server.
(eglot--TextDocumentItem): Finally, get correct language id.
* test/lisp/progmodes/eglot-tests.el (eglot--guessing-contact):
Enhance macro.
(eglot-test-server-programs-guess-lang): Update test.
Some editors have this. Depending on your preference, this can either
be wildly distracting and easily confused with actual code, or a
significant early aid that relieves you from moving around or reaching
for the mouse to consult an error message. To be safe, hide this
behind a customization variable and keep it disabled.
Personally, I find it less obstrusive and more helpful than expected.
* lisp/progmodes/flymake.el (flymake--delete-overlay): New helper.
(flymake--highlight-line): Handle flymake-show-diagnostics-at-end-of-line.
(flymake--clear-foreign-diags): Use flymake--delete-overlay.
(flymake--publish-diagnostics): Use flymake--delete-overlay.
(flymake-mode): Use flymake--delete-overlay.
(flymake-error-echo)
(flymake-warning-echo, flymake-note-echo): New faces.
(flymake-show-diagnostics-at-end-of-line): New option.
(Version): Bump to 1.3.4
* doc/misc/flymake.texi:
(Finding diagnostics): Mention flymake-show-diagnostics-at-end-of-line.
(Customizable variables): Mention
flymake-show-diagnostics-at-end-of-line and a few more relevant faces.
* etc/NEWS (Flymake): Mention flymake-show-diagnostics-at-end-of-line.
It turns out all calls to this function pass the same value
for the `count` argument, and they're all in `regexp.c`.
* src/syntax.c (RE_SETUP_SYNTAX_TABLE_FOR_OBJECT):
Rename from `SETUP_SYNTAX_TABLE_FOR_OBJECT`. Remove `count` argument.
Move call to `RE_SYNTAX_TABLE_BYTE_TO_CHAR` to where its result is
actually used.
* src/regex-emacs.c (re_search_2, re_match_2): Adjust accordingly.
* src/syntax.h (RE_SYNTAX_TABLE_BYTE_TO_CHAR): Rename from
`SYNTAX_TABLE_BYTE_TO_CHAR` to clarify that it takes a byteoffset as
used in the regexp engine and not a "bytepos". Adjust all callers.
That macro added 1 to buffer positions because:
Strings are 0-indexed, buffers are 1-indexed
but the reality is that this 1 was added to the regexp engine's "byte
offsets" which are not 1-based byte positions as used throughout
the rest of Emacs, but they are BEGV_BYTE-relative offsets, so the two
did not cancel out.
* src/regex-emacs.c (PTR_TO_OFFSET, POS_AS_IN_BUFFER): Delete macros;
use `POINTER_TO_OFFSET` instead.
(re_search_2, re_match_2, re_match_2_internal): Adjust accordingly.
* src/syntax.h (SYNTAX_TABLE_BYTE_TO_CHAR): Don't remove 1 from buffer
byteoffsets now that `POS_AS_IN_BUFFER` doesn't add it any more.
`gl_state` had an `offset` field because:
For buffers, regex-emacs.c passes arguments to the
UPDATE_SYNTAX_TABLE functions which are relative to BEGV
but the reality is that these arguments are byte offsets relative to
BEGV_BYTE whereas `offset` was counted in chars, so the two didn't
cancel each other out.
* src/syntax.h (struct gl_state_s): Delete `offset` field.
(UPDATE_SYNTAX_TABLE_FORWARD, UPDATE_SYNTAX_TABLE_BACKWARD)
(SYNTAX_TABLE_BYTE_TO_CHAR):
* src/syntax.c (SETUP_SYNTAX_TABLE, SETUP_SYNTAX_TABLE_FOR_OBJECT)
(update_syntax_table, skip_syntaxes): Simplify accordingly.
Also describe new 'echo-face' property in the Flymake manual, and fix
it's mistaken mention of a non-existing 'flymake-severity' property.
* doc/misc/flymake.texi:
(Flymake error types): Describe new properties and correct mention
of 'severity' property.
* lisp/progmodes/flymake.el:
(flymake-diagnostic-oneliner): Rework.
(flymake-error, flymake-warning, flymake-note): Add new 'echo-face' property.
(flymake--highlight-line)
(flymake-eldoc-function)
(flymake--tabulated-entries-1): Use flymake-diagnostic-oneliner
(Version): Bump to 1.3.3
* lisp/emacs-lisp/bytecomp.el (byte-compile-form)
(byte-compile-ignore):
Instead of compiling each `ignore` argument for value which is then
immediately discarded, compile it for effect but suppress
ignore-return-value warnings by passing the special value
`for-effect-no-warn` as for-effect parameter.
Proposed by Stefan Monnier.
* lisp/treesit.el (treesit-forward-sexp): Add docstring. Use
'restricted' tactic.
(treesit-transpose-sexps): Fix docstring.
(treesit-beginning-of-thing)
(treesit-end-of-thing): Add support for TACTIC.
(treesit-beginning-of-defun)
(treesit-end-of-defun): Supply treesit-defun-tactic as TACTIC.
(treesit--navigate-thing): Add support for TACTIC. Wrap the old form
in a new (if (eq tactic 'restricted) (new-code) (old-form)), and
supply the TACTIC parameter when recursing.
Only echo the first line of a potentially very large error message.
* lisp/progmodes/flymake.el:
(flymake-diagnostic-oneliner): New helper.
(flymake--tabulated-entries-1)
(flymake-eldoc-function): Use it.
(Version): Bump to 1.3.2.
(Package-Requires): Use Eldoc 1.14.0.
* lisp/gnus/nndiary.el (nndiary-last-occurrence):
Don't sort a list and throw away the result because that leaves the
list variable in an undefined state. Just take the largest element
because that was obviously what the author meant.
* lisp/allout.el (allout-process-exposed): Reverse the list of indices
properly so that the correct numbering for nodes in a subtree is used.
Avoid destructive reversing; the list may be a constant (literal).
This flaw was revealed by an ignored-return-value warning.
