Problem:
Tree-sitter uses UINT32_MAX for full-document ranges, which becomes -1 on 32-bit platforms and reaches _foldupdate as an invalid end row.
Solution:
Treat negative changed-range end rows as unbounded and clamp them to the buffer line count. Add a regression test that simulates the 32-bit sentinel.
AI-assisted: Codex
(cherry picked from commit 20ff82d9fc)
Problem:
urxvt doesn't support resetting to default cursor style like we assumed.
https://github.com/neovim/neovim/issues/38987#issuecomment-5004636437
Solution:
Instead of the reset sequence, always set it to steady block. This will
still be wrong for users who configured their initial cursor to
underline, but it's about the best we can do.
(cherry picked from commit 28818702e9)
Problem:
The code block query relied on a bug fixed in commit
7ed5609439, where a trailing newline was
not included in a node's text.
Solution:
Fix the query to expect the newline.
(cherry picked from commit fdc09be03c)
build: support variant builds via env-var and cpack overrides
Problem: scripts/genappimage.sh runs `make` with no extra flags and
emits a fixed nvim-linux-<arch>.appimage filename, so packaging a
variant build (e.g. with the bundled PUC-Rio Lua interpreter)
requires editing the Makefile or post-renaming the AppImage, which
breaks the AppImage zsync URL.
Solution: forward DEPS_CMAKE_FLAGS and CMAKE_EXTRA_FLAGS to `make`,
accept OUTPUT so the AppImage (and its zsync Filename:) follows the
caller's chosen name, and mark CPACK_PACKAGE_FILE_NAME as CACHE
STRING on Linux so the resulting tarball/deb follow as well. Default
behavior is unchanged when none of these env vars are set.
Problem: do_sub() only checks the timeout limit after finishing a line.
A pathological regex will run on a single line input unbounded
until the compute is completed.
Solution: Pass the timeout limit to `vim_regexec_multi()` so the
computation on the regex engine is bounded per-line.
Signed-off-by: XiaowenHu96 <me@xiaowenhu.com>
(cherry picked from commit bd73d8c011)
Problem:
on_attach() calls refresh(), but there is no guarantee the attached
buffer is the current buffer. This can make linked editing request
handling assume the wrong window.
Solution:
Call refresh() only if the current buffer is attached. This keeps the
initial highlighting behavior while avoiding making incorrect request.
(cherry picked from commit 9180c6e093)
Problem:
LSP hover erroneously drops blank lines before a 4-space-indented
codeblock, which is not valid Markdown. This causes incorrect parsing
and wrong display.
Solution:
Fix `split_lines` so that it doesn't drop the blank line just before
a 4-space-indented codeblock.
fix https://github.com/neovim/neovim/issues/40860
(cherry picked from commit 317c5ddda6)
fix(path): avoid shell fallback for unknown users
Problem:
Expanding an unknown ~user path falls back to the shell. In a fast event,
this re-enters the event loop and aborts Nvim.
Solution:
Leave unknown users unexpanded when the system account lookup fails.
(cherry picked from commit 9d3edf345f)
Co-authored-by: MAAZIZ Adel Ayoub <adelayoub.maaziz@gmail.com>
test: unreliable "put command . register special … ring the bell"
Problem:
FAILED …/put_spec.lua:894: …/put_spec.lua @ 898: put command . register special tests should ring the bell when deleting if not appropriate
…/put_spec.lua:894: …/put_spec.lua:898: Bell was not rung after action
Snapshot:
screen:expect([[
^ine of words 1 |
Line of words 2 |
{1:~ }|*4
{3:[No Name] [+] }|
|
{1:~ }|*4
{UNKNOWN_HL_ID(4):[No Name] }|
{UNKNOWN_HL_ID(1): }|
]])
stack traceback:
…/ui/screen.lua:917: in function '_wait'
…/ui/screen.lua:540: in function 'expect'
…/editor/put_spec.lua:894: in function 'bell_test'
…/editor/put_spec.lua:940: in function <…/editor/put_spec.lua:932>
Analysis:
The bell was asserted via the UI 'bell' event (screen.bell), but
vim_beep() rate-limits that to 3 per 500ms. These tests share one
long-lived session, so beeps from earlier tests could exhaust the window
and drop the asserted beep.
Solution:
Use assert_beeps()/assert_nobeep(), which check called_vim_beep (set
before the rate-limit gate).
Problem: #40731 may still crash if close_buffer autocmds reinsert the float's
grid. Plus removing the grid (and posting win_close) is unneeded if
win_close_othertab refuses to close the window later, which is possible.
Solution: do the stuff before freeing the window, like win_close.
(cherry picked from commit 1741da8412)
Problem: Closing a floating window from a non-current tab frees its grid
without removing it from the compositor's `layers` table, so the next
`ui_comp_put_grid()` walks a dangling pointer (UAF).
Solution: Call `ui_comp_remove_grid()` (and `ui_call_win_close()` for
multigrid UIs) before `win_free_mem()`, matching `win_close()` since
PR #21551.
(cherry picked from commit 6f370f34f2)
Problem
The current list of properties for inlay hint resolve support includes a
non-existent "location". This field should be "label.location" according
to the specification. Tooltips and commands for inlay hint "parts"
aren't advertised as being resolvable.
Solution
Remove "location" and add "label.location", "label.tooltip", and
"label.command".
Closes#40740
(cherry picked from commit 4bc30d6e10)
All wasm parsers shared a single global TSWasmStore via
ts_parser_set_wasm_store(). A TSWasmStore is owned by exactly one parser:
ts_parser_delete() frees it through ts_wasm_store_delete(). So the first
wasm parser collected freed the shared store, leaving the global pointer
dangling, and creating the next wasm parser dereferenced freed memory in
ts_wasm_store_reset() -> wasmtime_store_context(), crashing with SIGSEGV.
Give each parser its own store via ts_wasm_store_new() instead. This is
the 1:1 store-per-parser model intended by tree-sitter (see
tree-sitter/tree-sitter#3454): the global store remains only as the
language loader, ts_parser_delete() cleanly frees each parser's own
store, and the wasm engine is shared safely because ts_wasm_store_new()
clones its engine reference internally.
AI-assisted: Claude Code
Problem:
A `--listen` path longer than the system socket path limit (~104 bytes on
macOS, 108 on Linux) is silently truncated by `uv_pipe_bind()`. Nvim either
serves a socket at a different path than `v:servername` reports, or fails with
an error that blames the full untruncated path (confusing):
nvim: Failed to --listen: address already in use: "<full path>"
Steps to reproduce:
$ nvim --listen /var/folders/g7/9y_ydbnj2fs_fvp8xf44p8gc0000gn/T//nvim/-Users-rpatterson-Projects-src-github.com-neovide-neovide --embed -p
nvim: Failed to --listen: address already in use: "/var/folders/g7/9y_ydbnj2fs_fvp8xf44p8gc0000gn/T//nvim/-Users-rpatterson-Projects-src-github.com-neovide-neovide"
$ ls /var/folders/g7/9y_ydbnj2fs_fvp8xf44p8gc0000gn/T//nvim/-Users-rpatterson-Projects-src-github.com-neovide-neovide
"/var/folders/g7/9y_ydbnj2fs_fvp8xf44p8gc0000gn/T//nvim/-Users-rpatterson-Projects-src-github.com-neovide-neovide": No such file or directory (os error 2)
$ ls -l /var/folders/g7/9y_ydbnj2fs_fvp8xf44p8gc0000gn/T//nvim/
srwxr-xr-x@ - rpatterson 31 Mar 10:24 -Users-rpatterson-Projects-src-github.com-neovid
Solution:
Bind with `uv_pipe_bind2()` and `UV_PIPE_NO_TRUNCATE` (libuv 1.46+), so
a too-long path fails up front with the actual reason:
nvim: Failed to --listen: invalid argument: "<full path>"
(cherry picked from commit de141b4679)
Problem:
Clearing a register via `:let @a = ""` doesn't persist in shada.
Solution:
Follows the precedent of ee56daebb6 .
Namely, when the live register is empty and *at least as recent* as the copy
that is on disk, drop it instead of writing it back.
(cherry picked from commit aff618536b)
Co-authored-by: Barrett Ruth <62671086+barrettruth@users.noreply.github.com>
Problem:
`:let @/ = ""` doesn't stick after restart (the old search pattern comes back).
Solution:
When the current session cleared the pattern more-recently than stored
shada entry, drop the shada entry instead of restoring it.
Note: similar timestamp-based suppression was used for deleted
marks: ee56daebb6
(cherry picked from commit af9f24b80b)
Problem: :Man unloads hidden buffers, so reopening large pages reruns rendering
even when the width is unchanged.
Solution: keep regular man buffers loaded and refresh them only when the
effective width changes.
(cherry picked from commit a6827be98f)
Problem:
Session files specified at startup `-S [file]`, logically conflict
with `:restart`.
Solution:
Remove `-S [file]` from `v:argv` when doing :restart.
Also for the "bang" variant `:restart!`, just because it's
simpler (if anyone reports a use-case later, we can revisit).
(cherry picked from commit 47958bb4db)
Problem:
`:restart` does not preserve window layout, etc.
Solution:
- Change `:restart` to save/restore a session automatically.
- Introduce "bang" variant `:restart!` to restart *without* session
save/restore.
- Introduce `v:startreason`.
- `ZR` maps to `:restart!`.
(cherry picked from commit 845b66dd4a)
Problem:
- Transient mutation of the parent env is visible to any concurrent
code. Or at least just kinda sloppy.
- Latent bug:`channel_job_start` queues the spawn and returns before
`uv_spawn` runs, so the prior `os_unsetenv` immediately after the call
could in principle race with the deferred spawn.
Solution:
Pass `env` to the channel.
(cherry picked from commit a1da5d1f14)
Problem:
It's clumsy for scripts to handle a "restart", without custom mappings or
global vars.
Solution:
Introduce `v:startreason`
(cherry picked from commit ae426ee465)
The added test shows the context, one expected a JSON field
to be a Object but it was a null value
pros: shows `vim.NIL` instead of `a userdata`
cons: the context `field 'foo'` is lost. I think this is generated
with internal magic which is hard to replicate.
(cherry picked from commit 71b28da0f4)
Problem: On some systems `stderr` can be disabled. This results in not
usable `vim.pack` since it asserted `stderr` to be non-nil.
Solution: Stop asserting non-nil `stderr`. The downside is that
potential errors are not shown, but this is intentional since `stderr`
is disabled on system level.
Still assert non-nil `stdout` as its output is important for
`vim.pack` to actually do its job. Disabled `stdout` is not something
that can work with `vim.pack`.
(cherry picked from commit 25d33dd12b)
Problem:
When buffers are by default `nomodifiable`, such as when Nvim starts with
`-M`, the health buffer cannot be updated.
Solution:
Always set `modifiable` before modifying the buffer.
(cherry picked from commit 0f86ea5ed8)
fix(lsp): use LspNotify for document_color #40571
Problem: The document_color lsp module was already using the capability
framework but was still using raw buffer events to handle requests and
reloading. This means that every keystroke was sending a document_color
request to the server since there was no debounce in the raw handlers.
Solution: Switch to using LspNotify autocmd events. LspNotify fires just
after new document versions are synced with the server and provides a
built in debounce mechanism for changes. It also provides the signal for
when the current state should be cleared (didClose). The detach part is
already handled by the capability framework.
Fixes#39785
(cherry picked from commit 29db6ce84c)
Problem: Session with multiple tabpages sets 'winminheight' to 0.
Solution: Only save 'winminheight' and 'winminwidth' once (zeertzjq).
related: vim/vim#8119
related: neovim/neovim#40493closes: vim/vim#20673294dec827d
(cherry picked from commit c040f53dc1)
Problem: A previous refactor removed the BufWinEnter autocmd that
initiated a token request. When an LSP server sends a refresh
notification, then buffers that aren't shown in any window lost their
only trigger to request new tokens.
Solution: Add the BufWinEnter autocmd back which simply requests tokens
for all clients attached to the buffer.
Problem: changed_lines got a hardcoded 0, so the changelist entry
and '. mark always recorded column 0 instead of where the edit
actually happened.
Solution: pass start_col instead. changelist now tracks the real
column.
(cherry picked from commit 7312df77bb)
