Problem: UI callbacks disregard textlock unnecessarily (since d909de2).
Solution: In case the UI is flushed during textlock, make sure Nvim does
not flush UI callbacks that are expected to change text. Instead
assume postponing callbacks during textlock is insignificant,
and will safely happen soon after.
Problem:
Test sometimes fails on macos:
test/functional/terminal/channel_spec.lua:96: Expected objects to be the same.
Passed in:
(number) 3
Expected:
(number) 2
Solution:
Retry a few times.
Problem: When `vim._extui` is enabled, the default tabline shows an
incorrect window count (e.g., "2" when only 1 window is visible). This
happens because the extui pager window has `focusable=true` (needed for
user interaction) but `hide=true` (not visible by default).
Solution: Modify the window counting logic in `statusline.c` and
`ex_docmd.c:ex_tabs()` to also exclude hidden windows
(`wp->w_config.hide`), not just non-focusable windows.
Also updates test expectations in `cmdline2_spec.lua` which previously
expected the incorrect window count "2" when extui was enabled. The test
now correctly expects no window count indicator when only 1 visible
window exists.
Fixes#36759
Problem: Memory leak with :breakadd expr
Solution: Free debug_oldval and debug_newval before assigning to them.
Verify the existing (though confusing) :breakadd expr behavior
(zeertzjq).
It seems that :breakadd expr doesn't work as documented at all. This PR
only fixes the memory leak. The tests are for the existing behavior.
closes: vim/vim#18844a474de64df
Problem: Some tests are slow.
Solution: Make a few test cases faster.
bf63011a52
Only include the RunDbgCmd() to s:RunDbgCmd() change.
Co-authored-by: Bram Moolenaar <Bram@vim.org>
Problem: Debugger does not display the whole command.
Solution: Set ea.cmd before checking for a breakpoint.
33b55b562b
The code change is already covered by #35019.
Co-authored-by: Bram Moolenaar <Bram@vim.org>
Problem:
Sed thinks the argument starting with `1i` (and triple backticks) is the
suffix for the backup file because it comes right after the `-i` flag.
See for example
https://github.com/neovim/neovim/actions/runs/19774967693/job/56665991723.
Solution:
Explicitly mark it as command using the `-e` flag.
Problem:
MessagePack fixstr format supports string lengths 0-31, but mpack_str()
only used fixstr for lengths < 20. Strings of 20-31 bytes were
incorrectly encoded as str8 (2-byte header) instead of fixstr (1-byte
header).
Solution:
Change the condition from `len < 20` to `len < 32` to match the
MessagePack specification.
This fix affects message timing which exposed a pre-existing race
condition in the channels_spec PTY test. The test now uses a helper
function to accumulate partial PTY reads, making it more robust.
Fixes#32784
Allow the `request` parameter in `tokens_to_ranges` to be `nil` and
update version checking logic accordingly. This prevents errors when
the request is not present and improves robustness of semantic token
handling.
Problem: Cannot access the arguments of the current function in debug mode.
(Luc Hermitte)
Solution: use get_funccal(). (Lemonboy, closesvim/vim#1432, closesvim/vim#1352)
c7d9eacefa
This patch was regressed in e50b545676.
Co-authored-by: Bram Moolenaar <Bram@vim.org>
Problem: tests: excessive wait in Test_matchfuzzy_initialized
Solution: Use term_wait() instead of the TermWait() wrapper
(Corey Hickey)
Test_matchfuzzy_initialized is a terminal test, which are specified to be
"flaky" and automatically retried. The TermWait wrapper multiplies the
specified wait time by higher values for later retries, maxing out at
10x the specified value. This makes tries vim/vim#3 to vim/vim#6 sleep for 20 seconds
each, which makes the test very slow to work with. The specified intent
of the test (as noted in a comment eight lines above here) is to
sleep for 2s.
closes: vim/vim#18822d4f9de889b
Co-authored-by: Corey Hickey <bugfood-c@fatooh.org>
Problem: tests: Test_matchfuzzy_initialized() fails
Solution: Send a dummy key (Corey Hickey)
Test_matchfuzzy_initialized seems to expect that the 'lvimgrep' will be
interrupted by sending a SIGINT. If the search finishes beforehand,
though, then the SIGINT triggers vim to tell the user how to quit.
Vim does not show this message immediately, though; instead, vim shows
the message next time it is active. When StopVimInTerminal() sends a key
sequence intended to cause vim to quit, this activates vim to show the
message instead of quitting.
I do not understand every detail of the problem fully--if I type the
characters from StopVimInTerminal() into a post-SIGTERM terminal
directly, that seems to work ok; there seems to be a timing issue due to
sending all the characters at once.
This fix does make the test work reliably for me, and the test still
works even if I limit my CPU frequency so that the search is interrupted
by the SIGINT.
fixes: vim/vim#18821
related: vim/vim#18822a3925d783a
Co-authored-by: Corey Hickey <bugfood-c@fatooh.org>
Problem:
The `flags` field calls its sub-fields "experimental".
But `exit_timeout` is now used for multiple purposes.
Solution:
Graduate `exit_timeout` to a top-level ClientConfig field.
Problem: filetype: Erlang lexical files are not recognized
Solution: Detect *.xrl files as leex filetype, include syntax and
filetype plugins (Jon Parise).
leex is the lexical analyzer generator for Erlang. Its input file format
follows a section-based structure and uses the `.xrl` file extension.
This initial work includes file detection, an ftplugin (which inherits
the Erlang configuration), and a syntax definition.
Reference:
- https://www.erlang.org/doc/apps/parsetools/leex.html
related: vim/vim#18819closes: vim/vim#18832b087c5452b
Co-authored-by: Jon Parise <jon@indelible.org>
Problem: filetype: not all Erlang files are recognized
Solution: Detect *.app.src and rebar.config files as erlang filetype
(John Parise).
*.app.src files contain Erlang application definitions. (There are also
*.app files, which are similar but more often build artifacts, and that
file extension is too ambiguous to be recognized by default.)
Reference:
- https://www.erlang.org/doc/system/applications.html
Rebar is the Erlang build tool. rebar.config uses Erlang syntax.
Reference:
- https://rebar3.org/docs/configuration/configuration/closes: vim/vim#188352b2580e61a
Co-authored-by: Jon Parise <jon@indelible.org>
Problem:
Nvim supports `textDocument/semanticTokens/full` and `…/full/delta`
already, but most servers don't support `…/full/delta` and Nvim will try
to request and process full semantic tokens response on every buffer
change. Even though the request is debounced, there is noticeable lag if
the token response is large (in a big file).
Solution:
Support `textDocument/semanticTokens/range`, which requests semantic
tokens for visible screen only.
Problem:
Remote UIs can't execute lua code when a blocking prompt is waiting for
input. This is needed when implementing IME pre-edit for example.
Solution:
Add an `nvim__exec_lua_fast` experimental API function, which is allowed
to run instead of being queued until after the message has been shown.
Problem:
Strikethrough was not enabled for vtpcon thus preventing strikethrough
text from being shown.
Solution:
Update `windows.ti`.
Co-authored-by: ymich9963 <gmichael834@gmail.com>
Co-authored-by: Justin M. Keyes <justinkz@gmail.com>
Problem:
The rundll32 utility is a leftover from Windows 95, and it has
been deprecated since at least Windows Vista.
Solution:
Use the start command through Command Prompt instead.
Problem:
Currently we fetch a prebuilt xxd.exe from neovim/deps for Windows,
which is not ideal in terms of obviousness, misses updates, and is
clumsy to update.
Solution:
Similar to tee.c (#36363), vendor xxd.c from Vim and build it as part
of the Neovim build process. This makes the source obvious, enables
analyzer checks, and simplifies updates.
Fixes#36664
Problem: Returning 0 for has('patch-9.0.0') is inconsistent.
Solution: Make it return 1. (closesvim/vim#10640)
b0375d466e
Co-authored-by: Bram Moolenaar <Bram@vim.org>
Problem:
LSP server may not exit even after the client was stopped/disabled via enable(false).
Solution:
Automatically force-stop after a timeout, unless `client.flags.exit_timeout = false`.
Problem: has() is not strict about parsing the patch version.
Solution: Check the version more strictly. (Ken Takata, closesvim/vim#10752)
d90f91fe30
Co-authored-by: K.Takata <kentkt@csc.jp>
Problem: It is desirable for the default statusline to contain colored
diagnostics information. However, current `StatusLine` group is
purposefully defined as almost inverted `Normal` to "make current
window obvious". This makes diagnostic information from
`vim.diagnostic.status()` barely visible: it uses established
`DiagnosticSignXxx` groups which have colored foreground with
lightness close to `StatusLine` background.
Also the `StatusLineNC` group is fairly different from `Normal` in
order to both "makes window separators clear" and "be different from
`CursorLine`". But not as mush different as `StatusLine` because
"`StatusLine` and `StatusLineNC` should be clearly different".
Solution: Make both `StatusLine` and `StatusLineNC` be slightly closer
in lightness to `Normal`. This makes `StatusLine` and `StatusLineNC`
groups satisfy their conditions in the following way:
- `vim.diagnostic.count()` is readable on `StatusLine` - yes.
- `vim.diagnostic.count()` is readable on `StatusLineNC` - yes.
- `StatusLine` makes current window obvious - I'd say yes.
- `StatusLine` and `StatusLineNC` are clearly different - it depends
on the eyes and monitor. The current is clearly better, but the new
ones I'd say are still visibly different.
- `StatusLineNC` makes window separators clear - I'd say yes, but
depends on the eyes and monitor.
- `StatuslineNC` is different from `CursorLine` - NO, they are same.
Another approach to solve this would be to introduce dedicated
`DiagnosticStatuslineXxx` groups to use in `vim.diagnostics.status()`.
They can be defined using foreground colors from the same lightness as
`Normal`. This would make them readable in `StatusLine`. But not
`StatusLineNC`, though.
Problem:
Calling `:restart` twice erases the original args passed to `nvim`. This
is caused by interactions between the `:restart` command handler, the
`v:argv` parsing logic in the UI restart handler, and the options added
to `v:argv` by the server upon restart.
For example,
* Launch `nvim` as `nvim foo`:
* initial argv: `nvim foo`
* after nvim server launch: `nvim --embed foo`
* Run `:restart`
* after `ex_restart()`: `nvim -c '' --embed foo`
* after `remote_ui_restart()`: `nvim -c '' foo`
* after nvim server launch: `nvim --embed -c '' foo`
* Run `:restart` again
* after `ex_restart()`: `nvim -c '' --embed -c '' foo`
* after `remote_ui_restart()`: `nvim -c ''`
* after nvim server launch: `nvim --embed -c ''`
The intention of the argv parser in `remote_ui_restart()` is to only
take the first `-c cmd` and ignore any additional ones, but it actually
ignores the rest of argv when it encounters a second `-c` and there are
no `-` or `--` remaining.
Solution:
Fix the argv parser to reset the `skipping_minc` flag at the end of
every iteration that does not reach the `continue` statement.
