datetime.datetime.timestamp does not exist on Windows and
datetime.datetiem.strftime('%s') is not supported, since '%s' is a POSIX
format. Instead, use the recommended `calendar.timegm(obj.utctimetuple())`.
Problem: Dep3patch files are not recognized.
Solution: Recognize dep3patch files by their location and content. (James
McCoy, closesvim/vim#9367)
647ab4cede
PostgreSQL ships with man pages for SQL statements like `CREATE TABLE`,
which are provided with underscores as `man 7 CREATE_TABLE`. This patch
updates `man#open_page` (as used by `:Man`) such that visually selecting
the words `CREATE TABLE` in SQL code and pressing `K` properly opens the
desired man page.
Writing `:Man CREATE TABLE` still does not work, since `CREATE` is
interpreted as a section name. (Similarly, `:Man CREATE TABLE AS` fails
because there are too many arguments to `:Man`.) But this is okay,
because if you're typing it anyway then you can just enter underscores
and also tab-completion properly suggests `:Man CREATE_TABLE(7)`.
This is a bit bespoke, but my box has over 9000 man pages (as reported
by `man -k '' | wc -l`), and not one of them has a space in the man page
name, whereas the Postgres manuals do exist and are actually useful.
Test Plan:
On a machine with Postgres manual pages, running
nvim -u NORC +'exe "norm iCREATE TABLE foo(x int);" | norm 0veeK'
should open the appropriate man page.
wchargin-branch: man-spaces-to-underscores
Problem: Some plugins have structure `lua/nvim-someplugin/..`
Since `-` is not allowed in vim function names, healthcheck names in
lua and in vim can not have the same name (typically vim will use `_`
instead of `-`).
Solution: Normalize the names before checking for duplicates.
* fix(runtime/health): mitigate issues with duplicate healthchecks
Previously if a healthcheck was found as Lua and Vim it was executed
both times.
This new implementations prefers Lua, therefore if two are found It only
runs the Lua one, this way a plugin can mantain both implementations the
Lua one with the method `check()` and the autoload function `#check()`
(for none HEAD nvim versions).
**Note: This will require plugins to use `check()` as the function name,
since the autoload function that wraps the lua implementation won't be
called**
* docs(health): use spaces and don't overuse backtics
followup to #15259
- Refactor health.vim to discover lua healthcheck in the runtime
directories lua/**/health{/init}.lua
- Support healthchecks for lua submodules e.g :checkhealth vim.lsp and
also support wildcard "*" at the end for all submodules
:checkhealth vim*
- Refactor health.vim to use variable scope instead of output capturing
- Create health.lua module to wrap report functions and future
extensibility.
- Move away from searching just in the runtimepath, use
`nvim_get_runtime_file` due to #15632
Example:
Plugin linter in rtp can declare it's checkhealts in lua module
`lua/linter/health{/init}.lua` that returns a table with a method
"check" that when executed calls the report functions provided by the
builtin lua module require("health").
The plugin also has a submodule `/lua/linter/providers` in which it
defines `/lua/linter/providers/health{/init}.lua`
This plugin healthcheck can now be run by the ex command:
`:checkhealth linter linter.providers`
Also calling all submodules can be done by:
`:checkhealth linter*
And "linter" and "linter.provider" would be discovered when:
`:checkhealth`
Now remove the addition of "start/*" packages in 'packpath' as
explicit items in 'runtimepath'. This avoids 'runtimepath' from becoming
very long when using a lot of plugins as packages.
To get the effective search path as a list, use |nvim_list_runtime_paths()|
Problem:
`buftype=help` occasionally propagates from help to man buffer. As a result the
next time you open help it opens in the man window, replacing the manpage.
Test case:
nvim -u NORC
:Man man
:set bt? " should print `buftype=nofile`
:help
<C-W><C-W><C-W>c " go back to :Man window and close it
:help " focus help window
:Man man " open window with manpage again
:set bt? " prints `buftype=help`
Solution:
- call s:set_options()
- man#read_page() (called by autocmd BufReadCmd man://*) should already do
this. But BufReadCmd doesn't fire for already-existing man:// buffers.
Fix#15650
Problem:
"set filetype=man" assumes the user wants :Man features, this does extra
stuff like renaming the buffer as "man://".
Solution:
- old entrypoint was ":set filetype=man", but this is too presumptuous #15487
- make the entrypoints more explicit:
1. when the ":Man" command is run
2. when a "man://" buffer is opened
- remove the tricky b:man_sect checks in ftplugin/man.vim and syntax/man.vim
- MANPAGER is supported via ":Man!", as documented.
fixes#15487
Block copy and paste from system-clipboard currently breaks formatting.
This fixes it.
The bug occurs because system-clipboard doesn't contain information
about what mode the copy was made.
Simple solution to this is we keep a cache of copy we last made along
with mode information. If system-clipboard returns the cache we apply
the mode information that we know about that cache.
Update runtime files.
cb80aa2d53
Omit runtime/doc/tabpage.txt.
Patch v8.2.1401 is not ported yet.
Port optwin.vim changes without gettext().
Patch v8.2.1544 is not ported yet.
The current guidance for install Python packages is to use
python -m pip install <package_name>
Instead of
pip install <package_name>
This ensures that one is using the version of pip that is tied to the
environment's interpreter (and, thusly, its packages). This has [been
endorsed by a core
maintainer](https://snarky.ca/why-you-should-use-python-m-pip/) as
being the recommended way to invoke pip.
As there currently are a few places where the old invocation was used,
attempt to bring them in line.
Fixes#14234
- If the shada file is set with shada option n, use it.
- If the shadafile is NONE, it does not check for file read/write access.
- If the shada file does not exist, try to create it.
Here I use a negative number to decide whether the count has been
explicitly set. I think it unlikely that negative sections will ever be
created given that negative numbers complicate argument handling:
```
$ man -1 foo
man: invalid option -- '1'
```
and given that there's already precedence for alphanumeric sections like
`3p`, `3x`, `n`, etc.
---
This does work, though:
```
$ man -S -3 baz
```
With `man baz.-3` and `man 'baz(-3)'`, (GNU) man *might* consider `-3`
internally as a section, but in the end reports as if the whole
argument was the name of a topic:
```
$ man 'baz(-3)'
No manual entry for baz(-3)
```
---
Closes#13411.
In commit 63f0ca3263, `tagfunc` was introduced to
`runtime/autoload/man.vim`. Nonetheless the tag function instead
of using a short buffer name (e.g. `man://foo(3)`) uses the full
path to the man page (e.g. `man:///usr/share/.../foo.3.gz`). This
behaviour is inconsistent with `:Man!`, thus this commit.
Closes#13334
Problem: ruby#Detect() and node#Detect() don't return a [prog, err] pair
which means callers must special-case them.
Solution: align their return signatures with the perl/pythonx providers.
* support for :perl, :perlfile, :perldo and perleval()
* document that the perl provider doesn't currently work on Windows
* document that the perl legacy interface is now also supported
* added perleval() documentation
* import legacy perl interface tests
* only perl 5.22+ is supported
* healtcheck: use g:perl_host_prog if its set instead
using just 'perl' isn't correct as it may not be the version requested.
ditto for 'cpanm', rather go through 'App::cpanminus' to find the latest
perl version
using just 'perl' isn't correct as it may not be the version requested.
ditto for 'cpanm', rather go through 'App::cpanminus' to find the latest
perl version
I removed the SunOS stuff since no one uses SunOS and I've never tested
it on there.
I removed the section_flag init as we can just use -S instead of -s
and -S is used by every implementation as far as I know.
This brings man#init's time from 50-70ms to 15-20ms for me.
Closes#12318
Related #6766
Related #6815
Also, kudos to @zsugabubus for fixing a related issue in #12417
This also prevents any sorting of the paths from man. We need to
respect the order we get from it otherwise you end up loading
/Applications/Xcode.app/Contents/Developer/Platforms/MacOSX.platform/Developer/SDKs/MacOSX.sdk/usr/share/man/man1/ls.1
on MacOS instead of /usr/share/man/man1/ls.1
Also cleaned it up a little and made it faster.
Closes#9159 and #9271
Also changes man#extract_sect_and_name_ref to only return a single
section at a time. This fixes a bug in its usage in man#goto_tag
where get_paths would be called with multiple sections and it does
not support that.
I noticed that our tagfunc doesn't obey b:man_default_sects and
I'll fix that next.
When UV_OVERLAPPED_PIPE was used for the pipe passed to the child process, a
problem occurred with the standard input of the .Net Framework application
(#11809). Therefore, add the overlapped option to jobstart() and change it so
that it is set only when necessary
On some versions of Windows, WSL is unable to execute symbolic links to
Windows executables (microsoft/WSL#3999). As a workaround for that problem
this changes to use resolve() on WSL if win32yank was a symbolic link.
fixes#12113.
cpanm cannot look for Perl modules from root directories
without sudo so it creates '~/perl5/' and look for Perl modules in there.
Whether this directory existed before running cpanm or not,
cpanm returns a warning to advice the user to setup local::lib
in order to use modules in '~/perl5/' and exits with error code 0.
Each line in the warning always starts with '!'.
Display this warning to the user.
Continue parsing the version number if the warning can be ignored
because lines that are not prefixed with '!' are valid output.
Fix#11858
cpanm outputs a warning that suggest to use 'sudo' or use local::lib.
cpanm exits with 0 so nvim thinks that the command worked.
cpanm output that starts with "!" is likely an error.
Close#11858
fix#11753close#11781
The virtualenv troubleshooting in the Python provider health checks is
supposed to help the user determine whether running Python from Neovim
(as in `system('python')` or `system(exepath('python'))`) will use the
correct executable when a virtualenv is active. Currently however, it
issues spurious warnings in legitimate setups, and conversely, fails to
warn about potentially problematic ones.
See https://github.com/neovim/neovim/issues/11753#issuecomment-578715584
for a more detailed analysis, but at a high level, this is due to two
things:
- the virtualenv check is part of the Python provider check defined in
`s:check_python`, which uses a roundabout and sometimes erroneous way of
determining the Python executable
- more generally, it shouldn't be part of the provider check at all,
because it's not really related to the Python *provider*, i.e. the
Python executable which can communicate with Neovim via `pynvim`, but to
the Python the user is editing source files for, which typically
shouldn't even have `pynvim` installed
This patch reimplements the virtualenv check and factors it out into its
own separate function, which is however still kept in
`health/provider.vim` alongside the rest of the Python troubleshooting,
since troubleshooting all Python-related stuff in one place is probably
a good idea in order to alleviate any potential confusion (e.g. users
who run only provider checks might be left wondering whether their
virtualenv Python was properly detected if the report only shows their
global Python as the provider used by Neovim).
The old `:Man` implementation would take either the word under
the cursor, or the argument passed in, and load that as a man page.
Since we now use 'tagfunc' and look for all relevant man-pages, if
your system has several (i.e. same name, different sections), we return
several, giving the user an option.
This works for most tag commands except `:tjump`, which will
fail if there's multiple tags to choose from. This just happens to
be what the cscope code uses (it actually attempts to prompt the
user, but this fails).
test_registers.vim can fail even if a clipboard manager is running.
If a clipboard manager is not running, this test always fails with xclip.
Use xsel as a workaround.
https://github.com/astrand/xclip/issues/20 suggests closing stdout
when sending input via stdin.
Environment
- Ubuntu Xenial
- Vim 7.4 (any app with broken clipboard code will do)
- Neovim nightly
Steps to reproduce:
0. Start the clipboard manager.
1. Open a file in Vim on Linux.
Vim should have +clipboard enabled.
'set clipboard='
2. Yank some text to the clipboard register.
3. Quit Vim.
4. Run 'cd /path/to/neovim/repo/'
5. Run 'make oldtest'.
Do not run any individual tests.
They likely pass with or without this fix.
Before fix: test_registers.vim can fail.
After fix: test_registers.vim always passes.
Close https://github.com/neovim/neovim/issues/7958https://wiki.ubuntu.com/ClipboardPersistence#The_state_of_things
This addresses a minor quality problem with the recent `'tagfunc'`
changes for `man.vim` (see [link]).
Currently, with the cursor on a parenthese, hitting `K` will jump us to
the man page of the next mentioned entry, instead of the one to which
the parenthese (or section number) belongs.
```
pcrepattern(3), terminfo(5), glob(7), regex(7).
e.g. ^ e.g. ^
```
Adding the parentheses to `'iskeyword'` means we correctly handle these cases too.
[link]: https://github.com/neovim/neovim/pull/11280#discussion_r348342357
Mainly configuration and RPC infrastructure can be considered "done". Specific requests and their callbacks will be improved later (and also served by plugins). There are also some TODO:s for the client itself, like incremental updates.
Co-authored by at-tjdevries and at-h-michael, with many review/suggestion contributions.
Problem: If spellfile is missing, then "set spell" in modeline/sandbox
fails with a non-obvious error.
Solution: Check for sandbox early and give a meaningful error.
Fixes#11359
Test case:
# test.latex has `% vim: set spelllang=hu:`
# no spell file for `hu` yet!
nvim -u NORC --cmd 'autocmd FileType tex setlocal spell' --cmd 'set modeline' test.latex
Before:
No spell file for "hu" in utf-8
Download it?
Downloading hu.utf-8.spl...
Error detected while processing /usr/local/share/nvim/runtime/autoload/netrw.vim:
line 583:
E12: Command not allowed from exrc/vimrc in current dir or tag search: au WinEnter *^Iif &ft == "netrw"|call s:NetrwInsureWinVars()|endif
Error detected while processing function spellfile#LoadFile[60]..spellfile#Nread[13]..netrw#NetRead[4]..<SNR>67_NetrwOptionsSave:
line 66:
E171: Missing :endif
Error detected while processing function spellfile#LoadFile[60]..spellfile#Nread:
line 13:
E171: Missing :endif
Error detected while processing function spellfile#LoadFile:
line 60:
E171: Missing :endif
Error detected while processing modelines:
line 1:
E12: Command not allowed from exrc/vimrc in current dir or tag search
After:
Error detected while processing function spellfile#LoadFile:
line 5:
E605: Exception not caught: Cannot download spellfile in sandbox/modeline. Try ":set spell" from the cmdline.
Error detected while processing modelines:
line 1:
E12: Command not allowed from exrc/vimrc in current dir or tag search
sys.path.remove("") raises ValueError if the item is missing.
https://docs.python.org/3/library/functions.html#filter:
> filter(function, iterable) is equivalent to the generator expression (item
> for item in iterable if function(item))
fixes#11293
In order to find if there was already an open man page, the :Man command
would cycle through each window to see if &ft=='man'. This triggers
autocmds, e.g. BufEnter, unnecessarily and can have unexpected
side-effects.
Change the logic to check each window's ft without switching to it
unless it is actually a man window.
Signed-off-by: Joshua Rubin <me@jawa.dev>
The Python provider was special (via [1]), and would continue to do
checks with `0` being set explicitly even.
This was fixed in #11044 (45447e3b6), ref: #11040.
This extends it to use the same method with all providers.
1: https://github.com/neovim/neovim/pull/8047
This reverts part of ade88fe4c [1].
This is required for `let g:python3_host_prog = 'python'` etc, where it
should get picked up from PATH.
Without this it would show:
```
- INFO: pyenv: Path: /home/user/.pyenv/libexec/pyenv
- INFO: pyenv: Root: /home/user/.pyenv
- INFO: Using: g:python3_host_prog = "python"
- ERROR: "python" was not found.
- INFO: Executable: Not found
- ERROR: Detected pip upgrade failure: Python executable can import "pynvim" but not "neovim": python
- ADVICE:
- Use that Python version to reinstall "pynvim" and optionally "neovim".
pip3 uninstall pynvim neovim
pip3 install pynvim
pip3 install neovim # only if needed by third-party software
```
Note that it additionally causes a weird error
("Detected pip upgrade failure"), due to `s:check_bin` emptying
`python_exe` (because the non-absolute file not being readable), and
`provider#pythonx#DetectByModule('pynvim', a:version)` from 75593e6fce
then just getting the value from the host var again (without actual
checks).
This is implicitly fixed via this patch now (because it is skipped), but
could need some improvement in this regard probably.
With this patch it resolves it (for a virtualenv where pynvim is not
made available intentionally):
```
- INFO: pyenv: Path: /home/daniel/.pyenv/libexec/pyenv
- INFO: pyenv: Root: /home/daniel/.pyenv
- INFO: Using: g:python3_host_prog = "python"
- WARNING: $VIRTUAL_ENV exists but appears to be inactive. This could lead to unexpected results.
- ADVICE:
- If you are using Zsh, see: http://vi.stackexchange.com/a/7654
- INFO: Executable: /home/daniel/.pyenv/shims/tmp-system-deoplete.nvim-f205aF/python
- ERROR: Command error (job=11, exit code 1): `'/home/daniel/.pyenv/shims/tmp-system-deoplete.nvim-f205aF/python' -c 'import sys; sys.path.remove(""); import neovim; print(neovim.__file__)'` (in '/home/daniel/.dotfiles/vim/plugged/deoplete.nvim')
Output: Traceback (most recent call last): File "<string>", line 1, in <module>ModuleNotFoundError: No module named 'neovim'
Stderr: Traceback (most recent call last): File "<string>", line 1, in <module>ModuleNotFoundError: No module named 'neovim'
- INFO: Python version: 3.7.4
- INFO: pynvim version: unable to load neovim Python module
- ERROR: pynvim is not installed.
Error: unable to load neovim Python module
- ADVICE:
- Run in shell: pip3 install pynvim
```
Note: this appears to display the error twice via "Output:" and
"Stderr:".
1: https://github.com/neovim/neovim/pull/8784
Instead of deciding provider status in eval_has_provider, move the
decision to the provider Vim scripts.
Previously, provider loading worked as follows:
1. eval_has_provider() verified provider availability by searching for
the provider#providername#Call function and cached this verificaion as a static
variable for some providers
2. providers short-circuited on loading to prevent the definition of the
Call function (with the exception of the node provider that did not)
This commit changes the expected interface between nvim and its
providers to facilitate provider reloading, by splitting the
verification of the provider from the availability of the Call function.
eval_has_provider() now checks for a provider#providername#enabled
variable. It is up to the provider script to set this to 0 or 1
accordingly. eval_call_provider() remains unchanged.
All providers hosting a Call function were updated to respect this.
The clipboard provider now has a Reload function to reload the
provider.
`wl-copy` by default tries to determine the mime type of a copied bit of
text. From the [readme](https://github.com/bugaevc/wl-clipboard):
> wl-copy automatically infers the type of the copied content by running
> xdg-mime(1) on it.
So copying a Ruby script from Nvim may store it in the Wayland clipboard
as mime-type `application/x-ruby`.
This is a small reproduction without Nvim:
$ cat test.rb
#!/usr/bin/env ruby
puts 'hello world'
$ cat test.rb | wl-copy
$ wl-paste --list-types
application/x-ruby
This commit fixes that by telling wl-copy that all text copied from
Nvim has the mime type `text/plain`.
$ cat test.rb | wl-copy --type text/plain
$ wl-paste --list-types
text/plain;charset=utf-8
The neovim module is available for backwards compatibility. We should
not yet force the use of the pynvim module, since there's no other major
reason to bump the minimum supported Python client module.
Closes#9426
Before this commit, if user does this:
let g:node_host_prog = '~/.nvm/versions/node/v11.3.0/bin/neovim-node-host'
the "~/" is not expanded to user's home directory.
`:help g:ruby_host_prog` suggests a path with "~/" so technically we
already claimed to support this.
closes https://github.com/neovim/node-client/issues/102
PR #9304 added support for functions in clipboard providers. As part of
the PR I meant to move two checks in the provider code out of an if
statement into separate statements and adding additional checks for
g:clipboard attributes - as it turns out the code is wrong and it does
not implement additional checks while it adds two conditions that make
very little sense
type(g:clipboard['copy']) #isnot# v:t_func
what would make sense would be something along the lines of
type(g:clipboard['copy']['+']) #isnot# v:t_func
but might not be what we want either, so I'm reverting this.
Up to now g:clipboard["copy"] only supported string values invoked as
system commands.
This commit enables the use of VimL functions instead. The function
signatures are the same as in provider/clipboard.vim. A clipboard
provider is expected to store and return a list of lines (i.e. the text)
and a register type (as seen in setreg()).
cache_enabled is ignored if "copy" is provided by a VimL function.
The order was swapped in #4150 to prefer `xsel` but there wasn't a clear
explanation. Meanwhile, `xsel` has been neglected upstream.
Let's trying preferring `xclip` again, we've had a few reports of
problems with `xsel`.
closes#7237
ref #5853
ref #7449
Update runtime files.
fc65cabb15
---
vim-patch:8.0.1279: initializing menus can be slow
Problem: Initializing menus can be slow, especially when there are many
keymaps, color schemes, etc.
Solution: Do the globbing for runtime files lazlily. (Ken Takata)
man#init_pager() guesses the ref by the heading, which is usually
uppercase, so we don't know the correct casing. But lowercase is more
common, so use that for the buffer name instead of uppercase.
ref #9156
Before this commit, man#init_pager() always tries to scrape the manpage
name and set the buffer name. That's much less important than avoiding
duplicate buffers and E95. And it doesn't seem to be necessary, usually.
Steps to reproduce:
$ export MANPAGER="nvim -c 'set ft=man' -"
$ man sleep
:Man sleep
Error detected while processing function man#init_pager:
line 15:
E95: Buffer with this name already exists
:ls!
1 h- "man://SLEEP(1)" line 4
2 %a- "man://sleep(1)" line 1
For each package manager,
1. Check if they are executable before starting a job.
2. Check if the job opts are set before checking the job status
TODO: Validate the job id.
Prioritize npm because it's faster than yarn.
Reindent the code to 2-space indent.
When nroff justifies a line, it fills the line with whitespace to meet
$MANWIDTH. With $MANWIDTH=9999, that of course results in nonsense (and
behaves poorly with 'cursorline' option).
To work around that, instead of trying to hard-justify the lines, just
replace the mega-whitespace with a fixed size of 10 spaces.
Perhaps N/Vim needs a "soft justify" feature?
When rbenv is used for managing Ruby installations, just checking for an
executable called "neovim-ruby-host" is not enough. It has to be run as well.
If it does not return 0, then neovim-ruby-host is merely a shim for another Ruby
installation.
$ rbenv versions
2.5.0
* 2.5.1 (set by /Users/mhi/.rbenv/version)
$ rbenv whence neovim-ruby-host
2.5.0
$ which neovim-ruby-host
/Users/mhi/.rbenv/shims/neovim-ruby-host
$ neovim-ruby-host
rbenv: neovim-ruby-host: command not found
The `neovim-ruby-host' command exists in these Ruby versions:
2.5.0
$ echo $?
127
Additionally, the detection logic was moved from provider#ruby#Detect() to
s:detect(), because the former is run in the sandbox which forbids calling
system().
This would need to get `expand`ed to not become empty, and is being
handled by s:check_bin already.
`s:check_bin` will also complain about e.g.
"~/.pyenv/versions/3.6.6/bin/python" not being executable, but that
reflects that the host will fail to start with it.
Fixes#8778
Compare `$VIRTUAL_ENV` to `python_bin`.
This is necessary when `g:python_host_prog` is set to an absolute path,
and looking up `pyname` in `$PATH` yields another result.
OpenBSD's man returns all candidates when searching with -w instead of
the first one it finds. So this patch takes the first one if multiple
entries are found.
closes#8372closes#8341
Remove "" from sys.path (typically the first entry), which could cause
e.g. "logging" to be added from the current directory.
This gets done already for loading the host in
runtime/autoload/provider/pythonx.vim.
Problem: Loading file type detection slows down startup.
Solution: Move functions to an autoload script.
851ee6c3da
---
vim-patch:8.0.0635
Problem: When 'ignorecase' is set script detection is inaccurate.
Solution: Enforce matching case for text. (closes#1753)
I have `g:python3_host_prog` set to the system Python, where a package
is also installed to provide the "neovim" module.
`:checkhealth provider` however displays a warning for this:
> Your virtualenv is not set up optimally.
This is because /usr/bin/python is not in /home/user/.pyenv.
I think this warning should not get displayed if host_prog_var exists.
It goes back to the initial commit (20447ba09), and is maybe only
missing the `!` there as with the previous commit.
Full output:
```
- INFO: pyenv: /home/user/.pyenv/libexec/pyenv
- INFO: pyenv root: /home/user/.pyenv
- INFO: Using: g:python3_host_prog = "/usr/bin/python"
- WARNING: Your virtualenv is not set up optimally (/usr/bin/python is not in /home/user/.pyenv).
- ADVICE:
- Create a virtualenv specifically for Neovim and use `g:python3_host_prog`. This will avoid the need to install Neovim's Python module in each virtualenv.
- WARNING: $VIRTUAL_ENV exists but appears to be inactive. This could lead to unexpected results.
- ADVICE:
- If you are using Zsh, see: http://vi.stackexchange.com/a/7654
- INFO: Executable: /usr/bin/python
- INFO: Python3 version: 3.6.4
- INFO: python-neovim version: 0.2.1
- OK: Latest python-neovim is installed: 0.2.1
```
`g:loaded_python3_provider` gets set when the autoload file is sourced,
but this might error out, e.g. with deoplete:
[deoplete] Failed to load python3 host. You can try to see what happened by starting nvim with $NVIM_PYTHON_LOG_FILE set and opening the generated log file. Also, the host stderr is available in messages.
[deoplete] function remote#define#FunctionBootstrap[1]..remote#host#Require[10]..provider#pythonx#Require[13]..provider#Poll, line 14
[deoplete] deoplete requires Python3 support("+python3").
[deoplete] deoplete failed to load. Try the :UpdateRemotePlugins command and restart Neovim. See also :checkhealth.
It refers to `:checkhealth` from there explicitly, which would then
(without this patch) say that Python 3 is disabled.
This patch changes the reported info to include that it might have been
disabled due to some error, and keeps on going.
If `jobstart()` fails, then the subsequent `rpcrequest()` will throw due
to an invalid channel id. This causes `job.stderr` not to exist, so we
throw another exception when trying to dump the job's stderr.
Error detected while processing function remote#define#AutocmdBootstrap[1]..remote#host#Require[10]..provider#pythonx#Require:
line 22:
E716: Key not present in Dictionary: stderr
This obfuscates the actual problem.
On some versions of macOS, pbcopy doesn't work in tmux <2.6
https://superuser.com/q/231130
Fallback to tmux in that case.
Add a healthcheck for this scenario.
