When narrowing changes, parse reparses, so the timestamp should
definitely increment, just like in ts_record_changes.
Failing to increment this timestamp, outdated nodes would think they
are still up-to-date, and try to print their type name. Printing
their type name involves accessing the old parse tree, which is
already freed during the last reparse.
I also found that we don't increment timestamp when changing parser
ranges and fixed that as well.
* src/treesit.c (treesit_sync_visible_region):
(Ftreesit_parser_set_included_ranges): Increment timestamp.
* src/treesit.h (Lisp_TS_Parser): Add some comments.
* src/treesit.c (treesit_traverse_child_helper):
Do not call treesit_traverse_sibling_helper when the named node is
required and the last child is the named node.
Otherwise treesit_traverse_sibling_helper will move cursor to the
previous sibling and last node will be skipped.
* test/src/treesit-tests.el (treesit-search-subtree-forward-1):
(treesit-search-subtree-backward-1):
Add tests.
So turns out ts_node_field_name_for_child takes a named node index,
but we were passing it normal index that counts both named and
anonymous nodes. That's what makes the field name all wrong in
treesit explorer.
* doc/lispref/parsing.texi:
(Accessing Node Information): Update docstring.
* lisp/treesit.el (treesit-node-index): Add some unrelated comment.
(treesit-node-field-name): Get named node index rather than all node
index.
* src/treesit.c (Ftreesit_node_field_name_for_child): Update
docstring, use ts_node_named_child_count.
* src/xterm.c (x_frame_highlight, x_frame_unhighlight): Skip
changing the frame alpha when the frame is not eligible for
focus state-specific alpha values; otherwise, the alpha might be
reset by the time a alpha change wrought by a focus change
arrives, impeding handle_one_xevent from subsequently restoring
the initial value. (bug#66398)
* src/process.c (create_process) [F_SETPIPE_SZ]: Set the pipe
capacity only if the required read-process-max is larger than the
default capacity of the pipe. (Bug#66288)
* src/nsterm.m (ns_init_locale): If LANG is set, try to use that,
otherwise try to deduce what LANG should be. Check is the result is
valid, and use LANG="en_US.UTF-8" if not.
* doc/emacs/macos.texi (Mac / GNUstep Customization):
* src/nsfns.m (syms_of_nsfns): Fix documentation of
'ns-use-proxy-icon'; explain what it does and how it's
used. (Bug#66190)
Somehow the new overlay code in Emacs-29 changed slightly
the test to decide when to run the `modification-hooks` of overlays,
with the consequence that these hook functions end up being executed
also when text is deleted right after an empty overlay, which is
contrary to Emacs-28 behavior as well as contrary to the Texinfo doc.
* src/buffer.c (report_overlay_modification): Better reproduce the
Emacs-28 code.
* test/src/buffer-tests.el (overlay-modification-hooks):
Add corresponding test.
* src/treesit.c (treesit_cursor_helper_1)
(treesit_cursor_first_child_for_byte): Use
ts_tree_cursor_goto_first_child_for_byte first, and
ts_tree_cursor_goto_first_child second.
* src/pdumper.c (dump_buffer): Fix hash for 'struct buffer'. The
recent changes to that structure where commits
8f3091defb and
0bd4661941, both of which just affected
comments.
* doc/lispref/sequences.texi (Char-Tables):
* src/chartab.c (Fchar_table_range): Clarify what
'char-table-range' returns for an argument that is a cons cell.
* src/treesit.c: (treesit_cursor_helper_1): Use
ts_tree_cursor_goto_first_child_for_byte to speed up traversing among
siblings. The "while (ts_node_end_byte (cursor_node) < end_pos)" can
be removed with the check added in the loop below.
* src/w32term.c (w32_construct_mouse_wheel): The number of lines
to scroll should always be positive in wheel-scroll events.
Whether to scroll up or down is encoded in the modifiers, which
produce either wheel-up or wheel-down event. (Bug#65070)
* doc/lispref/commands.texi (Misc Events): Clarify the
documentation of 'wheel-up' and 'wheel-down' events.
* doc/emacs/search.texi (Replacement and Lax Matches):
* src/search.c (Freplace_match):
* lisp/replace.el (query-replace, query-replace-regexp): Clarify
in the doc string and the manual how letter-case is transferred
from the replaced text to the replacement text. (Bug#65347)
* src/xdisp.c (get_next_display_element): If we have no usable
face to display a character/composition, treat that as glyphless.
(produce_glyphless_glyph): If neither it->face nor its ASCII face
are usable, fall back to the frame's default font. (Bug#65198)
* src/emacs.c (main): Move the handling of the -x switch out of
the HAVE_X_WINDOWS condition, and simplify the rest of the code by
avoiding code duplication in HAVE_X_WINDOWS and !HAVE_X_WINDOWS
cases. (Bug#65048)
