With `overlap=true`, more extmarks than the requested limit may be
collected in `extmark_get`. This then leads to an out of bounds write of
`rv` in `nvim_buf_get_extmarks`.
Problem:
`:lsp restart` detects when a client has exited by using the `LspDetach`
autocommand. This works correctly in common cases, but breaks when
restarting a client which is not attached to any buffer. It also breaks
if a client is detached in between `:lsp restart` and the actual
stopping of the client.
Solution:
Move restart logic into `vim/lsp/client.lua`, so it can hook in to
`_on_exit()`. The public `on_exit` callback cannot be used for this, as
`:lsp restart` needs to ensure the restart only happens once, even if
the command is run multiple times on the same client.
Problem: The function get_logical_pos did not account for the possibility that a diagnostic might not have an associated extmark, leading to potential errors or incorrect behavior.
Solution: Add a check for diagnostic._extmark_id and return the logical positions directly if it does not exist.
Work on #37166
- Dynamic Registration Tracking via Provider
- Supports_Method
- Multiple Registrations
- RegistrationOptions may dictate support for a method
Problem:
We want to encourage implementing core features in Lua instead of C, but
it's clumsy because:
- Core Lua code (built into `nvim` so it is available even if VIMRUNTIME
is missing/invalid) requires manually updating CMakeLists.txt, or
stuffing it into `_editor.lua`.
- Core Lua modules are not organized similar to C modules, `_editor.lua`
is getting too big.
Solution:
- Introduce `_core/` where core Lua code can live. All Lua modules added
there will automatically be included as bytecode in the `nvim` binary.
- Move these core modules into `_core/*`:
```
_defaults.lua
_editor.lua
_options.lua
_system.lua
shared.lua
```
TODO:
- Move `_extui/ => _core/ui2/`
Problem:
Terminals should respond with the terminator (either BEL or ST) used in
the query so that clients can reliably parse the responses. The
`TermRequest` autocmd used to handle background color requests in the
terminal does not have access to the original sequence terminator, so it
always uses BEL. #37018
Solution:
Update vterm parsing to include the terminator type, then forward this
data into the emitted `TermRequest` events for OSC/DCS/APC sequences.
Update the foreground/background `TermRequest` callback to use the same
terminator as the original request.
Details:
I didn't add the terminator to the `TermResponse` event. However, I
assume the `TermResponse` event doesn't care about the terminator
because the sequence is already parsed. I also didn't update any of the
functions in `src/nvim/vterm/state.c` that write out responses. It
looked like those all pretty much used ST, and it would be a much larger
set of changes. In that same file, there's also logic for 8 bit ST
sequences, but from what I can tell, 8 bit doesn't really work (see `:h
xterm-8bit`), so I didn't use the 8 bit ST at all.
Problem: Deleting active plugins can lead to a situation when it is
reinstalled after restart.
Solution: Suggest "delete" code action only for not active plugins.
Whether a plugin is not active is visible by a hint in its header.
Problem: After `vim.pack.update()` it is not clear if plugin is active
or not. This can be useful to detect cases when plugin was removed
from 'init.lua' but there was no `vim.pack.del()`.
Solution: Add ` (not active)` suffix with distinctive highlighting to
header of plugins that are not active.
It will also be shown in in-process LSP document symbols to have quick
reference about which plugins are not active.
Problem: Using `vim.pack.del()` to delete active plugin can lead to
a situation when this plugin is reinstalled after restart. Removing
plugin from 'init.lua' is documented, but can be missed.
Solution: Make `del()` only remove non-active plugins by default and
throw an informative error if there is an active plugin.
Add a way to force delete any plugin by adding `opts.force`. This also
makes `del()` signature be the same as other functions, which is nice.
Problem: Buffer overflow in buf_write() when converting incomplete
multi-byte characters (Kevin Goodsell)
Solution: Make the buffer slightly larger
closes: vim/vim#19007f99de42a9f
Co-authored-by: Christian Brabandt <cb@256bit.org>
fix(pum): hide info window when insufficient space
Problem:
1. Info window was displayed even with insufficient space.
2. Tab characters counted as single cells.
Solution:
1. Hide window when space < 10 columns. Will be configurable
via completepopup width option in the future.
2. Use win_linetabsize over mb_string2cells.
Problem: There is now way to run `update()` without Internet connection
while there are some workflows that do not require it. Like "switch
plugin version" and "revert latest update".
Solution: Add `opts.offline` to `update()`. This also allows now to
treat `vim.pack.update(nil, { offline = true })` as a way to
interactively explore currently installed plugins.
Problem: There are two fairly common workflows that involve lockfile and
can be made more straightforward in `vim.pack`:
1. Revert latest update. Like if it introduced unwanted behavior.
2. Update to a revision in the lockfile. Like if already updated
on another machine, verified that everything works, `git add` +
`git commit` + `git push` the config, and want to have the same
plugin states on current machine.
Solution: Make `update` allow `opts.target`. By default it uses
`version` from a plugin specification (like a regular "get new changes
from source" workflow). But it also allows `"lockfile"` value to
indicate that target revision after update should be taken from the
current lockfile verbatim.
With this, the workflows are:
1. Revert (somehow) to the lockfile before the update, restart, and
`vim.pack.update({ 'plugin' }, { target = 'lockfile' })`. If Git
tracked, revert with `git checkout HEAD -- nvim-pack-lock.json`.
For non-VCS tracked lockfile, the revisions can be taken from the
log file. It would be nicer if `update()` would backup a lockfile
before doing an update, but that might require discussions.
2. `git pull` + `:restart` +
`vim.pack.update(nil, { target = 'lockfile' })`.
The only caveats are for new and deleted plugins:
- New plugins (not present locally but present in the lockfile)
will be installed at lockfile revision during restart.
- Deleted plugins (present locally but not present in the
lockfile) will still be present: both locally *and* in the
lockfile. They can be located by
`git diff -- nvim-pack-lock.json` and require manual
`vim.pack.del({ 'old-plugin1', 'old-plugin2' })`.
Problem:
When the popupmenu has no border but includes a scrollbar, the info window
column is misaligned due to a missing column offset.
Solution:
Apply a one-column offset to the info window when the popup menu has no border
and a scrollbar is present.
Problem: Changing `src` of already installed plugin currently takes
effect immediately inside `vim.pack.add()` and acts as "delete and
later fresh install". Although more robust, this might lead to
unintentional data loss (since plugin is deleted) if the plugin was
manually modified or the new source is not valid.
Also this introduces unnecessary differentiation between "change
`version`" and "change `src`" of already installed plugin.
Solution: Require an explicit `vim.pack.update()` to change plugin's
source. It is done by conditionally changing `origin` remote of the
Git repo. The effect does not require update confirmation in order to
have new changes fetched from the new `src` right away.
If in the future there are more types of plugins supported (i.e. not
only Git repos), also do extra work (like delete + install) during
`vim.pack.update()`.
Problem: patterns after a buffer-local pattern in a comma-separated aupat are
ignored.
Solution: ensure pat refers to the original buffer after each pattern, not the
buflocal_pat buffer, and when printing make sure it's normalized for
each pattern, not just the first.
Also simplify the logic when printing all autocommands for an event, and ensure
headings aren't repeated for each event when printing autocommands.
Background:
Suppose a window has concealed lines, and sets conceallevel>2,
concealcursor="". The concealed lines are displayed if the window is
curwin and the cursor is on the those lines.
Problem:
line('w$', win) switches curwin to win, and then does validate_botline
for curwin. It computes botline assuming the concealed lines displayed,
resulting in a smaller value than the actual botline that the user sees.
Solution:
Evaluate line('w$', win) without switching curwin.
Apply similar changes to other functions that switches curwin.
Co-authored-by: zeertzjq <zeertzjq@outlook.com>
Problem: error set by win_set_buf may leak if autocommands immediately close the
new window.
Solution: free the error set by win_set_buf. (prefer nvim_open_win's error as
it's more important and will cause 0 to be returned)
Problem: split_disallowed seemingly exists to prevent issues from changing
frames to accomodate a split window, which doesn't apply to floats.
Solution: remove the restriction for nvim_open_win, but only for floats.
(continue to check b_locked_split though)
NOTE: like before, the buffer we check b_locked_split for may not actually be
the target buffer "buf", as the later call to win_set_buf can fail to switch to
"buf" due to autocommands. (among other things)
Maybe we could attempt to close the new window in that case (or switch to a
different buffer if that also fails), but this is safer. (and simpler)
Fixes#36857 (and possibly some spurious E242s I've observed from extui)
Problem: vim.keymap.del has 'modes' as it's first argument while vim.keymap.set
has 'mode' as it's first argument despite both 'mode' and 'modes' taking in the
same type input of String or String[].
Solution: Updated vim.keymap.set docs to refer to it's first argument
as 'modes'.
Problem:
Tests that trigger `os_delay` messages may take 1-3 seconds, wasting
build/CI time, since this serves no purpose in tests.
Solution:
- Introduce `msg_delay` for cases where `os_delay` is being used as
a "UI feature".
- Skip `msg_delay` in tests.
Problem:
The q keymap is already set in open_floating_preview, so maparg('q') is not empty.
Solution:
Add a health.style check before setting the q keymap.
Problem: When fuzzy is enabled and the prefix is not empty,
items are not sorted by fuzzy score before calling fn.complete.
Solution: Use matchfuzzypos to get the scores and sort the items
by fuzzy score before calling fn.complete.
Problem:
CursorLine background check used global `normal_bg`, ignoring 'winhighlight'.
Solution:
Use `bg_attr` to get window-local Normal background instead.
Problem:
:messages history include the "search hit BOTTOM, continuing at TOP" message,
which is noise.
Solution:
Set msg_hist_off before giving the warning and reset it after warning.
Refactor capability checks in Client:_supports_registration and
Client:supports_method to properly handle dynamicRegistration and unknown
methods. Now, dynamic capabilities are checked before assuming support for
unknown methods, ensuring more accurate LSP feature detection.
Problem: feed_command, nvim_buf_set_option, nvim_buf_get_number, and exc_exec
are marked as deprecated.
Solution: Remove them from the test units in api/buffer_spec, autocmd/focus_spec,
ui/input_spec, and editor/put_spec.
Some test units only used a few deprecated functions, so creating a separate PR
for each would be excessive. Therefore, several were combined into a single PR.
