Problem:
Buffer-local CWD (:bcd) is "sticky", similar to window-local CWD (:lcd).
But this contradicts one of its main benefits: per-buffer "project root"
for LSP, OSC7.
Other problems:
- A buffer created with :edit/:enew/:new silently inherits b_localdir
(and b_prevdir) from the previous buffer.
- curbuf_reusable() refuses to recycle a scratch buffer that has
`b_localdir`.
- After :new/:vnew/:tabnew the CWD sticks to previous buffer's
`b_localdir` even though the new curbuf has none, so :new is not
equivalent to ":split | enew", and getcwd() disagrees with
haslocaldir().
- Requires "which buffer spawned this buffer" semantics that no other
buffer-local state has.
Solution:
Drop sticky/inherit behavior of buffer-local CWD (:bcd).
- do_ecmd: always apply the new curbuf's dir (`fix_current_dir`), like
`do_autochdir` already does. :tabnew from a :bcd buffer now reverts to
global CWD (and fires DirChanged), same as :tabnew from a :lcd window.
- curbuf_reusable(): recycling a scratch buffer frees its b_localdir.
To get sticky/inherit behavior of CWD, use `:lcd`.
Problem:
No way to set a buffer-local directory.
Use-cases:
- "Root dir" for LSP (and the "project" concept).
- `:terminal` OSC 7
Solution:
Add `:bcd` command.
- Extend `getcwd()` to take a third arg; `getcwd(-1, -1, bunfr)` returns
the buffer-local working directory.
- Buffer-local directories have less priority than window-local
ones, and more priority than tab-local ones.
Co-authored-by: Justin M. Keyes <justinkz@gmail.com>
Problem: :bwipe crashes if WinLeave wipes all other buffers
(after 9.1.2068).
Solution: Check for NULL pointer.
related: neovim/neovim#41066closes: vim/vim#208883e4019a082
Problem:
Cannot assign Lua functions/closures to "func" ('completefunc',
'tagfun', …) or "expr" ('foldexpr', 'indentexpr', …) options.
Solution:
- Store "func"/"expr" options as `Callback` instead of string.
- Delete oceans of copy-pasted code.
- BREAKING: LuaRef returned via RPC/Vimscript is now represented as
`"<Lua N: file:line>"` (like what `:map` shows) instead of `nil`.
- Note: `man.vim` still uses `v:lua` string, bc it's a vimscript ftplugin.
Helped-by: Lewis Russell <lewis6991@gmail.com>
Problem:
Options parsing is still painful for dict-style options.
Solution:
schema-maxxing => better `opt:get()` (will be the basis for `vim.o()`),
unified (and more-detailed) err msgs.
- Drop bespoke structure-builder in `_core/options.lua`.
- Define `schema` for all non-primitive options (except 'guicursor' and
statusline-style options); generate reified keysets `OptKeyDict`).
- Generate 'fillchars' => `fcs_tab`, 'listchars' => `lcs_tab`.
- `nvim_set_option_value`:
- Return the improved structures. Also from `vim.opt.x:get()`.
- Eliminate api <=> lua roundtrip, centralize option structure
handling.
- Improve/unify errors.
- Bump ERR_BUFLEN 80 → 256 so the "one of" list isn't truncated.
- Eliminate old 'diffopt' order-dependence (`iwhiteall` before `iwhite`)
Error samples:
Typed-key path (opt_strings_check → diffopt/mousescroll/breakindentopt):
E474: Unknown item 'foo'
E474: 'context' requires a number
E474: 'ver' number is out of range
E474: 'algorithm' must be one of: myers, minimal, patience, histogram
E474: 'filler' does not take a value
Related:
- #31084
- #34661
- #31820
- #14739
- #20107
- fix#18875
- :get() returns `{ sbr = true, shift = '3' }` (reified-keyset) instead of `{'sbr', 'shift:3'}`
- Setting via table now works too. `object_as_optval_for` `is_map` now recognizes struct options.
- fix#30296
- instead of `E474: Invalid argument`, errors now look like:
```
E474: Invalid value 'x', expected one of: single, double: ambiwidth=x
E474: Unknown item 'foo': diffopt=foo
E474: 'context' requires a number: diffopt=context:x
```
simplify `win_float_parse_option` from #26799.
Problem:
The preview-window (:pedit, etc.) always uses a split, but it would be
useful as a floatwin (or "popup").
Solution:
Support Vim's 'previewpopup' option.
Problem:
`FileType directory` fires *before* the `BufEnter` path where `dir.lua`
populates the buffer, in order to allow a 3P dir-browser plugin to handle the
event and (optionally) rename the buffer to e.g. `example:///tmp/foo/`.
But currently, `open_buffer()` continues after `filetypedetect BufRead` as if
the original directory buffer is *still* current and valid, which may fall
through to the built-in `nvim.dir` path after a 3P plugin ALREADY handled the
directory.
Solution:
Keep a buf ref around the early `filetypedetect BufRead` call for directory
buffers.
If the `FileType directory` handler switches away from/deletes the original
directory buffer, stop the open path instead of continuing into the built-in
`nvim.dir` flow.
Problem:
Continue separators normalization, and try to keep it at the
nvim <-> external boundary, e.g., `fn.xxx`, `api.xxx`, `:xxx`
Solution:
some key changes
- normalize `$HOME-windows`
- normalize entry points of changing directory
- `nvim_set_current_dir()`, `chdir()`
- normalize the named pipe
- `--listen` arg, `--server` arg, `:restart`
- `serverstart()`, `sockconnect()`, `serverstop()`
- make `expand()` respect 'shellslash' again
- clean up `did_set_shellslash`
- replace `forward_slash` with `TO_SLASH`
- remove obsolete `TMP_PATHSEPSTR`
- remove `slash_adjust` in `do_autocmd_dirchanged`?
- make `fnamemodify()` always return `/` (except when substituting
separators via `:s`, `:gs` ?)
Note:
- these funcs still apply `slash_adjust` on return, as before:
`:pwd`, `chdir()`, `exepath()`, `getcwd()`
- these funcs alwarys return `/`, unlike before the normalization PRs
- `getcompletion()`, `finddir()`, `findfile()`
Also
- clean up code, comments, formatting and tests
- move `TO_SLASH` from `f_chdir` to `changedir_func`
- move `TO_SLASH` from `f_bufadd` to `buflist_new`
Co-authored-by: Justin M. Keyes <justinkz@gmail.com>
Problem:
`getchar.c` is a quirky name and doesn't align with our existing
`os/input.c` and `tui/input.c` modules, which encompass the same topic
at different layers.
Solution:
Rename `getchar.c` => `input.c` for discoverability.
Introduce `dialog.c` for functionality related to showing confirm
prompts, dialogs, etc.
Problem:
`vim.filetype.match()` needs a cheap way to recognize directory buffers
without doing filesystem stat work.
Solution:
Ensure full buffer names for directories end in a trailing slash. Now
directory buffers can proceed through the normal 'filetype' path.
Note side-effects: session and ShaDa buffer-list restore behavior must
be compatible, so those + corresponding tests must be updated.
Problem:
cmdwin (the `:q` cmdline buffer) has various limitations which require
special-casing all over the codebase.
Besides complicating the code, it also breaks async plugins if they try
to create buffers/windows after some work is done, if the user happens
to open cmdwin at the wrong the moment:
Lua callback: …/guh.nvim/lua/guh/util.lua:531:
E11: Invalid in command-line window; <CR> executes, CTRL-C quits
stack traceback:
[C]: in function 'nvim_buf_delete'
…/guh.nvim/lua/guh/util.lua:531: in function <…/guh.nvim/lua/guh/util.lua:526>
Solution:
Just say no to "inception". Reimplement cmdwin as a normal buffer+window.
All of the cmdwin contortions (in both core, and innocent plugins) exist
literally only to support "inception": recursive
cmdwin-in-cmdline-things, like `<c-r>=`, `/`, search-during-substitute,
`:input()`, etc. So we just won't support that (though I have
a potential plan for that later, which I call "modal parking lot").
The benefit is that plugins, and core, no longer have to care about
cmdwin.
BONUS:
- mouse-drag on vertical separators works (it only worked for
horizontal/statusline before)
- inccommand-in-cmdwin now works correctly, for free (thus don't need
#40077).
POTENTIAL FOLLOWUPS
- Drop `CHECK_CMDWIN` ("E11: Invalid in command-line window"), allow chaos.
- Unify `BUFLOCK_OK` / `LOCK_OK` ?
DESIGN:
- Eliminate lots of C globals, `EX_CMDWIN`, etc.
- `text_locked()` no longer reports true for cmdwin.
- cmdwin = a normal window with 'winfixbuf', 'bufhidden=wipe',
'buftype=nofile'. Invariants come from those options rather than
special cases throughout the codebase.
- `nv_record` for q:/q//q? calls Lua
`nlua_call_vimfn("vim._core.cmdwin", …)`. No `K_CMDWIN`
/ cmdline-reader detour.
- `cedit_key` (`c_CTRL-F`) schedules a deferred event that calls
`vim._core.cmdwin.open(type, content, pos)` and returns `Ctrl_C` so
the in-flight cmdline cancels. Reader state is not serialized; instead
the captured `(type, line, col)` is replayed via
`nvim_feedkeys(type..line.."<CR>", "nt", …)` after user confirms.
- On confirm/cancel: `<CR>` / `<C-C>` calls into Lua which closes the
window and re-feeds the cmdline.
BREAKING CHANGES:
- Expression-register cmdline (`<C-R>=` from insert-mode) no longer
supports cmdwin. Same applies to `input()` / `inputlist()` (already
covered by `text_locked`).
- Usage of cmdwin in macros/mappings will probably break (assuming they
ever worked).
Problem: After maximizing and deleting the quickfix buffer, window
height is wrong (tertium)
Solution: Reset the winfixheight option when a quickfix buffer is
deleted from a window (Yegappan Lakshmanan)
fixes: vim/vim#3378closes: vim/vim#2040307f055f579
Co-authored-by: Yegappan Lakshmanan <yegappan@yahoo.com>
Problem:
Followup to #39858. close_windows returned a holder window just so
do_buffer_ext could retry there, but the non-float branch was dead code
and close_windows was called twice.
Solution:
Check one_window directly in do_buffer_ext, drop the holder, make
close_windows void again.
Co-authored-by: Sean Dewar <6256228+seandewar@users.noreply.github.com>
Problem:
(Followup to 54f22a8f01c0feb27a531b52aedf5cdbd5e51b24.)
Deleting another buffer from a floatwin could move focus into the holder
window and fire BufEnter for the buffer being deleted.
Solution:
Use switch_win_noblock() instead of buf_jump_open_win() before
recursing into do_buffer_ext().
Problem: home_replace() function can be improved
Solution: Refactor home_replace() to return the length of the string
(John Marriott).
In addition:
- in function set_b0_fname() move ulen into the block where it is used.
- In function findswapname() rework logic around displaying "swap file
already exists" dialogue so that literal message text is set once.
closes: vim/vim#20249a0931a90ee
Co-authored-by: John Marriott <basilisk@internode.on.net>
Problem:
`aucmd_restbuf` must be guarded in case `aucmd_prepbuf` wasn't called.
Solution:
Update `aucmd_restbuf` to be a no-op if `aucmd_prepbuf` wasn't called.
This requires `aco` to be zero-initialized.
Problem: with a float focused and the target buf only shown in the
last non-float window, do_buffer_ext goes down the buf != curbuf
path. close_windows can't touch the last non-float, b_nwindows stays
> 0, close_buffer is skipped, returns OK silently.
Solution: if a non-float still holds buf after close_windows, jump
into it and recurse. Then buf == curbuf and the existing replacement
path takes over.
Problem: Marks are not adjusted unloading a buffer that doesn't exist
on disk. E.g. extmarks are still valid (and will be beyond the
end of the buffer if the buffer is reloaded), even though the
text is lost.
Solution: Adjust marks for a cleared buffer when unloading a buffer that
doesn't exist on disk.
Problem: fileinfo not shown after :bd of last listed buffer
(memeplex)
Solution: Set need_fileinfo to true in empty_curbuf()
(Hirohito Higashi)
When deleting the last listed buffer with :bd, the new empty buffer's
file info (e.g. "[No Name]" --No lines in buffer--) was not displayed.
do_ecmd() only calls fileinfo() for existing buffers (oldbuf), not for
newly created empty buffers.
Set need_fileinfo in empty_curbuf() so the file info is displayed after
redraw.
fixes: vim/vim#548closes: vim/vim#198023d472d8675
Co-authored-by: Hirohito Higashi <h.east.727@gmail.com>
Co-authored-by: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
The NULL checks added in #35679 are no longer needed after #38473.
_____________________________________________________________________________________________
*** CID 645258: (REVERSE_INULL)
/src/nvim/buffer.c: 645 in close_buffer()
639 // Autocommands deleted the buffer.
640 emsg(_(e_auabort));
641 return false;
642 }
643 buf->b_locked--;
644 buf->b_locked_split--;
>>> CID 645258: (REVERSE_INULL)
>>> Null-checking "win" suggests that it may be null, but it has already been dereferenced on all paths leading to the check.
645 if (abort_if_last && win != NULL && one_window(win, NULL)) {
646 // Autocommands made this the only window.
647 emsg(_(e_auabort));
648 return false;
649 }
650 }
/src/nvim/buffer.c: 626 in close_buffer()
620 // Autocommands deleted the buffer.
621 emsg(_(e_auabort));
622 return false;
623 }
624 buf->b_locked--;
625 buf->b_locked_split--;
>>> CID 645258: (REVERSE_INULL)
>>> Null-checking "win" suggests that it may be null, but it has already been dereferenced on all paths leading to the check.
626 if (abort_if_last && win != NULL && one_window(win, NULL)) {
627 // Autocommands made this the only window.
628 emsg(_(e_auabort));
629 return false;
630 }
631
Problem: w_locked can be bypassed when recursively set if not restored
to its prior value.
Solution: Rather than save/restore everywhere, just make it a count,
like other locks (Sean Dewar)
Requires the previous commit, otherwise b_nwindows will be wrong in
tests, which causes a bunch of weird failures.
closes: vim/vim#197287cb43f286e
Co-authored-by: Sean Dewar <6256228+seandewar@users.noreply.github.com>
Problem: close_buffer() callers incorrectly handle b_nwindows,
especially after nasty autocmds, allowing it to go
out-of-sync. May lead to buffers that can't be unloaded, or
buffers that are prematurely freed whilst displayed.
Solution: Modify close_buffer() and review its callers; let them
decrement b_nwindows if it didn't unload the buffer. Remove
some now unneeded workarounds like 8.2.2354, 9.1.0143,
9.1.0764, which didn't always work (Sean Dewar)
(endless yapping omitted)
related: vim/vim#19728bf21df1c7b
b_nwindows = 0 change for free_all_mem() was already ported.
Originally Nvim returned true when b_nwindows was decremented before the end was
reached (to better indicate the decrement). That's not needed anymore, so just
return true only at the end, like Vim. (retval isn't used anywhere now anyways)
Set textlock for dict watchers at the end of close_buffer() to prevent them from
switching windows, as that can leave a window with a NULL buffer. (possible
before this PR, but the new assert catches it; added a test)
Despite textlock, things still aren't ideal, as watchers may observe the buffer
as unloaded and hidden (b_nwindows was decremented), yet still in a window...
Likewise, for Nvim, wipe_qf_buffer()'s comment may not be entirely accurate;
autocmds are blocked, but on_detach callbacks (textlocked) and dict watchers may
still run. Might be problematic, but those aren't new issues.
Co-authored-by: Sean Dewar <6256228+seandewar@users.noreply.github.com>
Problem: if close_buffer() in set_curbuf() unloads curbuf, NULL pointer
accesses may occur from enter_buffer() calling
end_visual_mode(), as curbuf is already abandoned and possibly
unloaded. Also, selection registers may not contain the
selection with clipboard+=autoselect(plus).
Solution: Move close_buffer()'s end_visual_mode() call to buf_freeall(), after
any autocmds that may restart it, but just before freeing anything
(Sean Dewar)
related: vim/vim#19728a8fdfd4fcb
Maybe this should be considered partial? clipboard+=autoselect isn't
implemented. If it is, we may need to update these comments to mention
TextYankPost being possible, and unskip its test.
Co-authored-by: Sean Dewar <6256228+seandewar@users.noreply.github.com>
Problem:
Currently, we recommend always inserting text above prompt-line in
prompt-buffer. This can be done using the `:` mark. However, although
we recommend it this way it can sometimes get confusing how to do it
best.
Solution:
Provide an api to append text to prompt buffer. This is a common
use-case for things using prompt-buffer.
Problem:
- Small error in port of v9.1.0678, causing :ball to check w_locked for the
wrong window.
- After #27439, free_all_mem() may not wipe out buffers that were open in more
than one window before windows were freed.
Solution:
- Check win_locked() for wp in ex_buffer_all(), not curwin.
- Set b_nwindows to 0 in free_all_mem() before calling close_buffer().
Ref: https://github.com/neovim/neovim/pull/38473#issuecomment-4125117681
No need to block these fixes on that.
free_all_mem() change also looks like it fixed the existing "N lua references were
leaked!" warnings on the CI.
Problem: 'jumpoptions' "view" doesn't remember skipcol and may lead to
glitched display with 'smoothscroll'.
Solution: Save skipcol in the mark view. Also make sure skipcol doesn't
exceed line size.
Problem: 5943a81 skips saving window options in `buflist_altfpos` for
style=minimal windows, which also prevents the window from restoring its own
options when switching buffers.
Solution: revert the change. In `get_winopts`, don't restore options from the
`WinInfo` of style=minimal windows when reusing values for a different window.
In `win_free`, clear `wi_optset` for minimal windows.
Problem: closing a minimal float saves its style-imposed options into
buffer's wininfo, which get picked up by normal windows on :bnext.
Solution: don't save window options to wininfo for style=minimal windows.
Problem:
Currently, if prompt gets changed during user-input with
prompt_setprompt() it only gets reflected in next prompt. And that
behavior is not also consistent. If user re-enters insert mode then the
user input gets discarded and a new prompt gets created with the new
prompt.
Solution:
Handle prompt_setprompt eagerly. Update the prompt display, preserve user input.
Problem: if buf_free_all autocommands open a terminal, it will remain open after
the buffer is freed.
Solution: close terminals again later, this time while blocking autocommands.
Did consider terminal_open checking stuff like b_locked_split instead, but
that's set during BufHidden, etc., which doesn't mean the buffer's being wiped.
Problem: Newlines intended to write messages below the cmdline or to
mark the start of a new message on message grid are emitted
through ext_messages. This results in unnecessary newlines for
a UI that has decoupled its message area from the cmdline.
msg_col is set directly in some places which is not transmitted
to msg_show events.
Various missing message kind for list commands.
Trailing newlines on various list commands.
Solution: Only emit such newlines without ext_messages enabled.
Use msg_advance() instead of setting msg_col directly.
Assign them the "list_cmd" kind.
Ensure no trailing newline is printed.
Problem: :tab sbuffer may close old tabpage if BufLeave autocommand
splits window (after 9.1.0143).
Solution: Only close other windows if the buffer will be unloaded
(zeertzjq).
related: neovim/neovim#37749closes: vim/vim#193526da9f757c4
Problem: Crash with combination of terminal popup and autocmd.
Solution: Disallow closing a popup that is the current window. Add a check
that the current buffer is valid. (closesvim/vim#5754)
cee52204ca
Cherry-pick related changes from patch 9.0.1454.
Co-authored-by: Bram Moolenaar <Bram@vim.org>
Problem: Last buffer not freed with EXITFREE (after 9.1.2087).
Solution: Free the last buffer when inside free_all_mem()
(zeertzjq).
This isn't really a memory leak, as the last buffer's memory is still
reachable via pointers like firstbuf and lastbuf. But it's possible that
this may cause false ASAN warnings in the future, which is what EXITFREE
is supposed to prevent.
closes: vim/vim#191946c118afeaa
Problem: Crash when using :tabonly in BufUnload.
Solution: Set curbuf when setting curwin->w_buffer. Don't wipe out a
buffer if there are no other buffers. Don't decrement
b_nwindows if it was 0 before buf_freeall() (zeertzjq).
fixes: vim/vim#19088#issuecomment-3710172769
closes: vim/vim#19186fa64f92f6a
Removed code doesn't seem to do anything? Looks like a clobbered remnant from
when do_filetype_autocmd lived in did_set_string_option.
Doc comment for wipe_buffer doesn't decrement top_file_num since a2d25b7 (2016),
which presumably means the comment on marks doesn't apply either. (fmark_T::fnum
can't refer to the wrong buffer as numbers aren't reused here anymore)
Problem: A BufAdd autocommand may cause alist_add() to use freed
memory, this is caused by the w_locked variable unset too
early (henices)
Solution: in trigger_undo_ftplugin() only set w_locked to false, if it
was false when calling the function.
related: v9.1.0678
closes: vim/vim#190239266a2a197
Co-authored-by: Christian Brabandt <cb@256bit.org>
Problem: b:undo_ftplugin not executed when re-using buffer
(archy3)
Solution: explicitly execute b:undo_ftplugin in buflist_new() when
re-using the current buffer
fixes: vim/vim#17113closes: vim/vim#17133baa8c90cc0
Cherry-pick test_filetype.vim changes from patch 9.1.1325.
Co-authored-by: Christian Brabandt <cb@256bit.org>
Problem: :bdelete/bunload/bwipeout may attempt to switch to a closing
buffer, which fails. (after 9.1.2058)
Solution: don't consider switching to closing buffers (Sean Dewar)
closes: vim/vim#1910763d53de72d
Co-authored-by: Sean Dewar <6256228+seandewar@users.noreply.github.com>
Problem: b_locked_split is not checked for :sbuffer, which allows
autocommands to leave windows open to freed buffers.
Solution: In do_buffer_ext, check just before possibly splitting, after
handling 'switchbuf'. Leave win_split to handle the check for
curbuf. (needed even if curbuf is not the target, as setting
the buffer after splitting may fail) (Sean Dewar)
closes: vim/vim#19096ac5c8ab6cc
Co-authored-by: Sean Dewar <6256228+seandewar@users.noreply.github.com>
