Problem:
vim.uv.os_setenv gets "stuck" per-key. #32550
Caused by the internal `envmap` cache. #7920
:echo $FOO <-- prints nothing
:lua vim.uv.os_setenv("FOO", "bar")
:echo $FOO <-- prints bar (as expected)
:lua vim.uv.os_setenv("FOO", "fizz")
:echo $FOO <-- prints bar, still (not expected. Should be "fizz")
:lua vim.uv.os_unsetenv("FOO")
:echo $FOO <-- prints bar, still (not expected. Should be nothing)
:lua vim.uv.os_setenv("FOO", "buzz")
:echo $FOO <-- prints bar, still (not expected. Should be "buzz")
Solution:
- Remove the `envmap` cache.
- Callers to `os_getenv` must free the result.
- Update all call-sites.
- Introduce `os_getenv_noalloc`.
- Extend `os_env_exists()` the `nonempty` parameter.
Problem: Message kind logic for emitting an error message is convoluted
and still results in emitting an unfinished message earlier than
wanted.
Solution: Ensure emsg_multiline() always sets the kind wanted by the caller
and doesn't isn't unset to logic for emitting the source message.
Caller is responsible for making sure multiple message chunks are
not emitted as multiple events by setting `msg_ext_skip_flush`...
Problem:
`nvim -es` (and `nvim -Es`) is the recommended way to non-interactively
run commands/vimscript. But it enables shada by default, which is
usually not wanted.
Solution:
- Disable shada by default for `nvim -es/-Es`. This can be overridden by
`-i foo` if needed.
- Do NOT change the 'loadplugins' default.
- User config + packages _should_ be enabled by default, for both `nvim
-es` and `nvim -l`. Else any Lua packages you have can't be accessed
without `-u path/to/config`, which is clumsy.
- Use-cases:
```
nvim --headless "+Lazy! sync" +qa
would become: nvim -es "+Lazy! sync"
nvim --headless +PlugInstall +qall
would become: nvim -es +PlugInstall
```
- Opt-out (`--clean` or `-u NONE`) is much easier than opt-in (`-u
path/to/config`).
- User config/packages are analogous to pip packages, which are
expected when doing `python -c ...`.
related: 7c94bcd2d7
related: ddd0eb6f51
Problem:
Can't use `n.clear()` to test non-RPC `nvim` invocations. So tests end
up creating ad-hoc wrappers around `system()` or `jobstart()`.
Solution:
- Introduce `n.spawn_wait()`
- TODO (followup PR): Rename `n.spawn()` and `n.spawn_wait()`.
It's misleading that `n.spawn()` returns a RPC session...
Problem: Option metadata like list of valid values for an option and
option flags are not listed in the `options.lua` file and are instead
manually defined in C, which means option metadata is split between
several places.
Solution: Put metadata such as list of valid values for an option and
option flags in `options.lua`, and autogenerate the corresponding C
variables and enums.
Supersedes #28659
Co-authored-by: glepnir <glephunter@gmail.com>
The newer UCRT runtime has native support for UTF-8, including forcing
it as the active codepage even before `main()` is called. This means
the c runtime will properly convert windows WCHAR:s into UTF-8 bytes,
as early as the argv/argc params to `main()` . Whereas MSVCRT does not
support this reliably and required us to use `wmain()`.
Only MSVC supports using manifest files directly as source files.
The solution for other Windows toolchains is to use a .rc file.
Problem: testing of options can be further improved
Solution: split the generated option test into test_options_all.vim,
add more test cases, save and restore values, fix use-after-free
closes: vim/vim#158946eca04e9f1
Co-authored-by: Milly <milly.ca@gmail.com>
In the api_info() output:
:new|put =map(filter(api_info().functions, '!has_key(v:val,''deprecated_since'')'), 'v:val')
...
{'return_type': 'ArrayOf(Integer, 2)', 'name': 'nvim_win_get_position', 'method': v:true, 'parameters': [['Window', 'window']], 'since': 1}
The `ArrayOf(Integer, 2)` return type didn't break clients when we added
it, which is evidence that clients don't use the `return_type` field,
thus renaming Dictionary => Dict in api_info() is not (in practice)
a breaking change.
Problem:
The default builtin UI client does not declare its client info. This
reduces discoverability and makes it difficult for plugins to identify
the UI.
Solution:
- Call nvim_set_client_info after attaching, as recommended by `:help dev-ui`.
- Also set the "pid" field.
- Also change `ui_active()` to return a count. Not directly relevant to
this commit, but will be useful later.
Problem:
- "process" is often used as a verb (`multiqueue_process_events`), which
is ambiguous for cases where it's used as a topic.
- The documented naming convention for processes is "proc".
- `:help dev-name-common`
- Shorter is better, when it doesn't harm readability or
discoverability.
Solution:
Rename "process" => "proc" in all C symbols and module names.
Problem:
$XDG_RUNTIME_DIR may be broken on WSL, which prevents starting (and even
building) Nvim. #30282
Solution:
- When startup fails, mention the servername in the error message.
- If an autogenerated server address fails, log an error and continue
with an empty `v:servername`. It's only fatal if a user provides a bad
`--listen` or `$NVIM_LISTEN_ADDRESS` address.
Before:
$ nvim --headless --listen ./hello.sock
nvim: Failed to --listen: "address already in use"
$ NVIM_LISTEN_ADDRESS='./hello.sock' ./build/bin/nvim --headless
nvim: Failed to --listen: "address already in use"
After:
$ nvim --headless --listen ./hello.sock
nvim: Failed to --listen: address already in use: "./hello.sock"
$ NVIM_LISTEN_ADDRESS='./hello.sock' ./build/bin/nvim --headless
nvim: Failed $NVIM_LISTEN_ADDRESS: address already in use: "./hello.sock"
Problem:
`nvim --listen` does not error on EADDRINUSE. #30123
Solution:
Now that `$NVIM_LISTEN_ADDRESS` is deprecated and input *only* (instead
of the old, ambiguous situation where it was both an input *and* an
output), we can be fail fast instead of trying to "recover". This
reverts the "recovery" behavior of
704ba4151e, but that was basically
a workaround for the fragility of `$NVIM_LISTEN_ADDRESS`.
This also makes shada reading slightly faster due to avoiding
some copying and allocation.
Use keysets to drive decoding of msgpack maps for shada entries.
Followup to #28515:
Rename the static os_homedir() to os_uv_homedir() to emphasize that it
is a wrapper around a libuv function.
Add the function os_get_homedir() to os/env.c to return the cached
homedir value as a const. Must be called after homedir is initialized or
it fails.
The difference between this function and the static os_uv_homedir() is
that the latter gets the homedir from libuv and is used to initialize
homedir in init_homedir(), while os_get_homedir() just returns homedir
as a const if it's initialized and is public.
Use the os_get_homedir() accessor for ~/ expansion on Windows to make
the code more concise.
Add a Windows section to main_spec.lua with tests for expanding ~/ and
~\ prefixes for files passed in on the command-line.
Signed-off-by: Rafael Kitover <rkitover@gmail.com>
Since paste data is handled via a separate channel, the data processed via `input_buffer` is typically just explicit keys as typed in by the user. Therefore it should be fine to use `memmove()` to always put the remaining data in front when refilling the buffer.
FileDescriptor is used to buffer togheter many small writes to fewer
syscalls. if the data to write already is in a single buffer, it is
perfectly fine to just use os_write directly (which will take care of
the reverse problem: splitting a too big write into many syscalls)
In command_line_scan() for MSWIN, expand "~\" or "~/" prefixed paths to
the USERPROFILE environment variable for the user's profile directory.
Fix#23901
Signed-off-by: Rafael Kitover <rkitover@gmail.com>
Instead of painfully messing with timing to determine if queries were
reparsed, we can simply keep a counter next to the call to ts_query_new
Also memoization had a hidden dependency on the garbage collection of
the the key, a hash value which never is kept around in memory. this was
done intentionally as the hash does not capture all relevant state for the
query (external included files) even if actual query objects still
would be reachable in memory. To make the test fully deterministic in
CI, we explicitly control GC.
In particular, TUI manages its own screen buffers and highlight table, so we don't need
to run init_highlight() and default_grid_alloc() in the ui client process.
Problem: `set_string_option_direct()` contains a separate codepath specifically for setting string options. Not only is that unnecessary code duplication, but it's also limited to only string options.
Solution: Replace `set_string_option_direct()` with `set_option_direct()` which calls `set_option()` under the hood. This reduces code duplication and allows directly setting an option of any type.
Problem Description: In ex_mode, the default_grid.chars are not allocated, and subsequently,
the w_grid.target in curwin is not allocated to default_grid in update_screen. This leads to
a null pointer crash when the completion function is executed in ex_mode.
Solution: Set full_screen when in ex_mode to ensure that default_grid is allocated.
Before, we needed to always pack an entire msgpack_rpc Object to
a continous memory buffer before sending it out to a channel.
But this is generally wasteful. it is better to just flush
whatever is in the buffer and then continue packing to a new buffer.
This is also done for the UI event packer where there are some extra logic
to "finish" of an existing batch of nevents/ncalls. This doesn't really
stop us from flushing the buffer, just that we need to update the state
machine accordingly so the next call to prepare_call() always will
start with a new event (even though the buffer might contain overflow
data from a large event).
Then we can just load metadata in C as a single msgpack blob. Which also
can be used directly as binarly data, instead of first unpacking all the
functions and ui_events metadata to immediately pack it again, which was
a bit of a silly walk (and one extra usecase of `msgpack_rpc_from_object`
which will get yak shaved in the next PR)
Functions like file_open_new() and file_open_fd_new() which just is a
wrapper around the real functions but with an extra xmalloc/xfree around
is an anti-pattern. If the caller really needs to allocate a
FileDescriptor as a heap object, it can do that directly.
FileDescriptor by itself is pretty much a pointer, or rather two:
the OS fd index and a pointer to a buffer. So most of the time an extra
pointer layer is just wasteful.
In the case of scriptin[curscript] in getchar.c, curscript used
to mean in practice:
N+1 open scripts when curscript>0
zero or one open scripts when curscript==0
Which means scriptin[0] had to be compared to NULL to disambiguate the
curscript=0 case.
Instead, use curscript==-1 to mean that are no script,
then all pointer comparisons dissappear and we can just use an array of
structs without extra pointers.
Note: this contains two _temporary_ changes which can be reverted
once the Arena vs no-Arena distinction in API wrappers has been removed.
Both nlua_push_Object and object_to_vim_take_luaref() has been changed
to take the object argument as a pointer. This is not going to be
necessary once these are only used with arena (or not at all) allocated
Objects.
The object_to_vim() variant which leaves luaref untouched might need to
stay for a little longer.
and for return value of nlua_exec/nlua_call_ref, as this uses
the same family of functions.
NB: the handling of luaref:s is a bit of a mess.
add api_luarefs_free_XX functions as a stop-gap as refactoring
luarefs is a can of worms for another PR:s.
as a minor feature/bug-fix, nvim_buf_call and nvim_win_call now preserves
arbitrary return values.
Problem:
Since 2448816956, the --startuptime report shows
two blocks of data. The TUI process and its embedded nvim process write to the
file concurrently, which may interleave the two startup sequences into the same
timeline.
Solution:
Report each process as a separate section in the same file.
1. Each process buffers the full report.
2. After startup is finished, the buffer is flushed (appended) to the file.
Fix#23036
Sample report:
--- Startup times for process: Primary/TUI ---
times in msec
clock self+sourced self: sourced script
clock elapsed: other lines
000.006 000.006: --- NVIM STARTING ---
000.428 000.422: event init
000.728 000.301: early init
...
005.880 000.713: init highlight
005.882 000.002: --- NVIM STARTED ---
--- Startup times for process: Embedded ---
times in msec
clock self+sourced self: sourced script
clock elapsed: other lines
000.006 000.006: --- NVIM STARTING ---
000.409 000.403: event init
000.557 000.148: early init
000.633 000.077: locale set
...
014.383 000.430: first screen update
014.387 000.003: --- NVIM STARTED ---
Remove `export` pramgas from defs headers as it causes IWYU to believe
that the definitions from the defs headers comes from main header, which
is not what we really want.
Problem: Many places in the code use `findoption()` to access an option using its name, even if the option index is available. This is very slow because it requires looping through the options array over and over.
Solution: Use option index instead of name wherever possible. Also introduce an `OptIndex` enum which contains the index for every option as enum constants, this eliminates the need to pass static option names as strings.
