Problem:
fnamemodify with the :r flag will not strip extensions if the filename
starts with a ".". This means that files named ".in" could cause an
infinite loop.
Solution:
Add early return if the filename was not changed
(cherry picked from commit 032e024f8a)
fix(vim.fs): dirname() returns "." on mingw/msys2 #30480
Problem:
`vim.fs.dirname([[C:\User\XXX\AppData\Local]])` returns "." on
mingw/msys2.
Solution:
- Check for "mingw" when deciding `iswin`.
- Use `has("win32")` where possible, it works in "fast" contexts since
b02eeb6a72.
Problem: When input immediately follows end of bracketed paste, the
nvim_input may be processed before the nvim_paste.
Solution: Ensure some waiting after the end of a bracketed paste.
(cherry picked from commit 5acdc4499e)
Problem:
Noise in stderr from src/nvim/po/CMakeLists.txt during build:
48 substitutions on 48 lines
"optwin.js" "optwin.js" 1299L, 57582B written
385 substitutions on 385 lines
"nvim.pot" "nvim.pot" 9318L, 180177B written
Solution:
Run with `-c silent source` instead of `-S`.
(cherry picked from commit cac86e9b4c)
There appears to be a bug in `make` where if a rule asks `make` to
invoke a command called `foo`, and `foo` exists somewhere in `$PATH` as
a directory (not an executable file), `make` will attempt to `execve`
that directory instead of continuing to search in later parts of the
`$PATH` for `foo` as a true executable.
The cause can be traced back to a bug in Make 4.3 which stems from their
use of the findprog function in Gnulib. This was reported to the Make
maintainers here:
https://savannah.gnu.org/bugs/index.php?57962
and then forwarded to the Gnulib maintainers here:
https://github.com/coreutils/gnulib/commit/7b1de4a
Make 4.4 does not have this bug, and I can confirm that I'm able to run
make in the Neovim repo with no further modifications to my system than
upgrading the version of make I'm using to 4.4 or 4.4.1.
As the change is small enough, and it's unlikely that make version
around the world is going to be updated in a timely manner, it makes
sense to just add a workaround for this. Using `command -v` to resolve
the `cmake` command, similar to what is already being done with
`cmake3`, makes it work correctly in all cases. Continuing to include
`... || echo cmake` at the end means that if neither `cmake3` nor
`cmake` are installed, the user will still see a message about CMake
being missing.
Co-authored-by: Jake Zimmerman <zimmerman.jake@gmail.com>
(cherry picked from commit e268fcbdaa)
This will fix the following error when using generators that have a
space in them, e.g. "Unix Makefiles":
"CMake Error: Could not create named generator Unix".
Closes https://github.com/neovim/neovim/issues/30218.
(cherry picked from commit ef8067a19d)
Using powershell as the default windows shell as using cmd alters $PATH
in a way that makes building neovim fail (powershell prioritizes visual
studio tools which is arguably more correct).
This was tested with gnu make for windows, which can be installed with
e.g. scoop. It does not work with nmake and it is extremely unlikely we
want to add nmake support as the makefile is merely supposed to be
syntactic sugar for the most common case. For similar reasons, the only
supported generator is ninja.
(cherry picked from commit 4ee65484b1)
This reverts commit 7f6ff829aa.
The given reasoning and usecase is unsatisfactory. While it is true that
it allows to run `make build/bin/nvim`, it can easily be recreated with
`ninja -C build bin/nvim` which does the exact same thing. This minor
convenience is not worth adding the extra code given how rare this
usecase should be.
(cherry picked from commit 3085c9d9da)
**Problem:** Top-level anonymous nodes are not being checked by the
query linter
**Solution:** Check them by adding them to the top-level query
This commit also moves a table construction out of the match iterator so
it is run less frequently.
This will ensure automatic backports created by the backport action does
not request reviewers (since the commit in question has already been
vetted and merged), but manual backports created by users does request
reviewers as these commits has not been vetted previously.
(cherry picked from commit 755512ed60)
Problem: Crash with cursor-screenline and narrow window
(elig0n)
Solution: Don't set right_col when width2 is 0 (zeertzjq).
fixes: vim/vim#15677closes: vim/vim#1567859149f0269
(cherry picked from commit 5191a11d66)
Problem: Wrong cursor-screenline when resizing window
Solution: Invalidate saved left_col and right_col when width1 or width2
change.
closes: vim/vim#1567986dc4f8b43
(cherry picked from commit 90585e47fe)
Problem: heap-use-after-free in garbage collection with location list
user data.
Solution: Mark user data as in use when no other window is referencing
the location list (zeertzjq)
fixes: neovim/neovim#30371closes: vim/vim#15683be4bd189d2
Problem:
The LSP omnifunc can insert nil bytes, which when read in other places
(like semantic token) could cause an error:
semantic_tokens.lua:304: Vim:E976: Using a Blob as a String
Solution:
Use `#line` instead of `vim.fn.strlen(line)`. Both return UTF-8 bytes
but the latter can't handle nil bytes.
Completion candidates can currently insert nil bytes, if other parts of
Alternative fix to https://github.com/neovim/neovim/pull/30359
Note that https://github.com/neovim/neovim/pull/30315 will avoid the
insertion of nil bytes from the LSP omnifunc, but the change of this PR
can more easily be backported.
(cherry picked from commit 8512f669f0)
`path2name()` function doesn't process `'pluginname/health/init.lua'` correctly. Instead of retruning `'pluginname'` it returns `'pluginname.health'`
(cherry picked from commit 237d2aef4d)
Problem:
str_utfindex_enc could return an error if the index was longer than the
line length. This was handled in each of the calls to it individually
Solution:
* Fix the call at the source level so that if the index is higher than
the line length, utf length is returned
(cherry picked from commit f279d1ae33)
Problem:
str_byteindex_enc could return an error if the index was longer than the
lline length. This was handled in each of the calls to it individually
Solution:
* Fix the call at the source level so that if the index is higher than
the line length, line length is returned as per LSP specification
* Remove pcalls on str_byteindex_enc calls. No longer needed now that
str_byteindex_enc has a bounds check.
while at it, also move the note about :wincmd
directly to :h :wincmd, it doesn't seem to belong to the buffer section.
closes: vim/vim#15636b584117b05
Co-authored-by: Christian Brabandt <cb@256bit.org>
(cherry picked from commit b40ec083ae)
Problem: Marktree meta count still includes invalidated marks, making
guards that check the meta total ineffective.
Solution: Revise marktree metadata when in/revalidating a mark.
Problem:
crash when calling nvim_buf_get_text() with a large negative start_col:
call nvim_buf_get_text(0, 0, -123456789, 0, 0, {})
Solution:
clamp start_col after subtracting it from the line length.
(cherry picked from commit d1d7d54680)
Problem: [security]: invalid cursor position may cause a crash
(after v9.1.0038)
Solution: Set cursor to the last character in a line, if it would
otherwise point to beyond the line; no tests added, as it
is unclear how to reproduce this.
Github Advisory:
https://github.com/vim/vim/security/advisories/GHSA-4ghr-c62x-cqfh396fd1ec29
Co-authored-by: Christian Brabandt <cb@256bit.org>
(cherry picked from commit 6f167fcae9)
