fix(api): auto-load buffers in nvim_buf_set_* operations
Problem: Setting marks, lines, or text on unloaded buffers fails because
ml_line_count is not accurate until the buffer is loaded.
Solution: Add require_loaded_buffer() helper and use it in write operations
(nvim_buf_set_lines, nvim_buf_set_text, nvim_buf_set_mark). These now
auto-load buffers as needed.
Problem: has("terminfo") test fails on local runs because it expects
"HAVE_UNIBILIUM " in the :version info to mean that nvim was built
without unibilium.
Solution: Assume nvim is built with unibilium if HAVE_UNIBILIUM is
present in the version string and is NOT followed by 0, false, or off.
This isn't affecting CI runs because when the test detects that it is
running in CI it doesn't use the :version string at all.
Resolves https://github.com/neovim/neovim/issues/37456
* test(core/channels_spec): fix flaky test
Always use expect_stdout() to check PTY output.
* test(autocmd/termxx_spec): fix flaky test
Usually the Ctrl-C cancels the following :qa!, but sometimes it doesn't,
and Nvim exits before feed() returns. Instead make sure that :qa! always
reaches Nvim and use expect_exit().
Problem:
Crashed Nvim leaves socket files. Subsequent starts fail with "address already in use".
Solution:
On bind failure, test socket liveness via connect(). Remove if stale, retry bind().
Fixes#36581
Problem:
With shadow border, scrollbar and border overlap. Shadow's right border
starts from the second row (offset by 1), but scrollbar renders on every
row including the first. If they share a column, the first row shows
scrollbar where shadow hasn't started yet, causing misalignment.
Solution:
Separate scrollbar from the shadow's right border.
Problem: A message ending in an unprintable character may emit a
msg_show event with the "empty" kind. No empty message event
for echom "".
Solution: Adjust conditions for emitting "empty" msg_show events.
Problem:
Creating an autocommand which executes `:bwipe` on the Syntax event
causes a heap-use-after-free.
Solution: set BF_SYN_SET flag before applying autocommands
Problem:
When text is too long, it overflows into the border. This happens because
grid_line_maxcol includes border columns, so grid_line_puts can write
there.
Solution:
Truncate string to available width when need_fcs_trunc is set.
When width1 and width2 are negative the assertion may fail. It seems
that adding a negative value to w_curswant won't cause any problems, so
just change the assertion.
Problem: nvim_parse_expression null pointer dereference with unmatched ]
followed by a node.
Solution: if ast_stack was empty, set new_top_node_p to top of the stack after
pushing the list literal node; similar to what's done for curlies.
This bug was originally found by a Matrix user, but I couldn't remember how to
trigger it... Ran into the other crash while finding a repro. :P
Problem: nvim_parse_expression null pointer dereference when parsing an
identifier followed by { with "highlight" parameter set to false.
Solution: only set opening_hl_idx if pstate->colors is not NULL.
Not added to parser_tests.lua as that uses highlight = true.
Problem: Error message for layout change does not match action.
Solution: Pass the command to where the error is given. (closesvim/vim#11573)
9fda81515b
Thinking about this again, it's actually OK to check split_disallowed in
window_layout_locked(), so add the check.
Also add missing window_layout_locked() in tabpage_close().
Co-authored-by: Bram Moolenaar <Bram@vim.org>
Problem: nvim_get_option_value with "filetype" set can crash if autocommands
open the dummy buffer in more windows, or if &bufhidden == "wipe".
Solution: Attempt to close all dummy buffer windows before wiping. Promote the
dummy buffer to a normal buffer if that fails.
Problem: When the "filetype" key is set for nvim_get_option_value, autocommands
can crash Nvim by prematurely wiping the dummy buffer, or cause options intended
for it to instead be set for unrelated buffers if switched during OptionSet.
Solution: Don't crash. Also quash side-effects from setting the buffer options.
Problem: Spell navigation skips words on the first line because
_on_spell_nav passes an empty range (0,0) to the highlighter.
Solution: Use math.max(erow, srow + 1) to ensure a valid search window.
Signed-off-by: ashab-k <ashabkhan2000@gmail.com>
Problem:
`t.skip()` adds to the "pending" list. If there is no path to fixing
a pending test, it adds noise to the pending list.
Solution:
Return early instead of using `t.skip()`.
Problem: Use-after-free in winframe_remove() (henices)
Solution: Set window_layout_locked() inside winframe_remove()
and check that writing diff files is disallowed when the
window layout is locked.
It can happen with a custom diff expression when removing a window:
1. Buffer was removed, so win_frame_remove() is called to remove the
window.
2. win_frame_remove() → frame_new_height() → scroll_to_fraction()
→ diff_check_fill() (checks for filler lines)
3. diff_check_fill() ends up causing a diff_try_update, and because we
are not using internal diff, it has to first write the file to a
buffer using buf_write()
4. buf_write() is called for a buffer that is not contained within a
window, so it first calls aucmd_prepbuf() to create a new temporary
window before writing the buffer and then later calls
aucmd_restbuf(), which restores the previous window layout, calling
winframe_remove() again, which will free the window/frame structure,
eventually freeing stuff that will still be accessed at step 2.
closes: vim/vim#19064ead1dda74a
Nvim doesn't have this bug as Nvim uses a floating window as autocommand
window, and removing it doesn't need winframe_remove().
Co-authored-by: Christian Brabandt <cb@256bit.org>
On linux /dev/stdin is defined as a symlink to /proc/self/fd/0
This in turn is defined as a "magic" symlink which is allowed to point
to internal kernel objects which really does not have a file
name. As a glaring inconsistency, fopen("/proc/self/fd/0", "r")
works if fd was originally opened using pipe() but not using
socketpair(). As it happens UV_CREATE_PIPE does not create pipes
but creates socket pairs. These two unfortunate conditions
means that using /dev/stdin and similar does not work in
shell commands in nvim on linux. as a work around, override
libuv's descicion and create an actual pipe pair.
This change is not needed on BSD:s but done unconditionally for simplicity,
except for on windows where it is not done for stdout because of windows
fixes#35984
Problem: Installing plugin with submodules doesn't check out their
state (due to `git clone --no-checkout` to not end up with default
branch code in case of invalid `version`).
Updating a plugin with submodules doesn't update their state.
Solution: Update `git_checkout` helper to account for submodules.
Another approach would be `git checkout --recurse-submodules ...`,
but that doesn't seem to allow `--filter=blob:none` for submodules,
which is nice to have.
Also make `git_clone` wrapper simpler since `--no-checkout` makes
`--recurse-submodules` and `--also-filter-submodules` do nothing.
Problem: The `add` startup tests mock startup process which requires
*some* amount of time to be done. Previous solution with `vim.wait`
to wait just enough time to register that startup script has finished
doesn't seem to work as intended (it just wait full time). Its timeout
is also seems to be barely enough to pass on Windows CI. Which will be
a problem with more `git` actions done on startup in the future/next
commit.
Solution: Just sleep predetermined amount of time and explicitly check
if startup script finished executing.
Problem:
vim.glob.to_lpeg() errors when patterns contain numeric literals
(like the '1' in '.ps*1') because LPeg interprets numeric strings
as indexed grammar rule references. For example:
vim.glob.to_lpeg('.ps*1')
E5108: Lua: rule '1' undefined in given grammar
Solution:
Prefix all rule names with '_' in the end_seg() function to prevent
literal numbers from being interpreted as LPeg indexed rules. This
ensures pattern components like '1', '2', etc. are treated as
regular rule names rather than special references.
Problem: Installing plugin is done via `git clone --no-checkout ...`
(to not end up with default branch code in case of invalid `version`).
This leaves cloned repo in a state that `git stash` will actually add
an entry to the stash list. Although not critical, better to not have
that if possible.
Solution: explicitly skip `git stash` step in checkout during install.
Problem:
vim.lsp.tagfunc looks for the presence of 'c' (cursor) flag and issues
sync textDocument/definition requests to all clients, otherwise
workspace/symbol requests. But 'c' flag can also be set during the
insert mode completion, e.g. with an empty tag completion query, the tag
func receives pattern of '\<\k\k' with flags 'cir'.
Solution:
check for 'i' (insert mode completion) flag and don't issue any LSP
requests, return vim.NIL for immediate fallback to tags.
This fixes a regression from cf6f60ce4d
(possibly), as before that commit a frame is popped from the call stack
immediately after its response is received.
Also fix leaking the allocated error messages.
Problem: info/preview floating windows are created
without wrap enabled, causing info text to be
truncated.
Solution: enable 'wrap' and set 'scrolloff' to 0 by
default, like vim's popup_create does.
Problem:
Using nvim_set_current_win() to switch windows while in Visual mode
causes E315 ml_get error when target buffer has fewer lines. This
doesn't happen with `:wincmd w` since it properly resets Visual mode
when switching buffers.
Solution:
Reset Visual mode when switching to another buffer, like `:wincmd w`.
Problem: when shortmess doesn't have 'S', backward search wrap doesn't
show the "W" before count. forward search works fine but
backward fails because the position check logic is backwards -
it checks if cursor < pos instead of using the existing
wrapped flag.
Solution: Use sia->sa_wrapped flag that searchit() already sets
correctly (glepnir).
fixes: vim/vim#5280closes: vim/vim#19138ccb7b43365
Co-authored-by: glepnir <glephunter@gmail.com>
Problem:
The `"Running healthchecks..."` message doesn't inform the user much and
is a hack from before we got a way to emit actual progress messages.
Solution:
Use `nvim_echo` to emit progress messages showing the name of the report
that is currently running.
Problem: Setting pumborder=shadow doesn't blend proerly.
Solution: Check fconfig.shadow when setting pum_grid.blending, like what
is done in check_blending().
