It is enabled with ENABLE_ASAN_UBSAN like other compilers. Technically
it only enables ASAN as UBSAN is not available, meaning to make the
variable names fully correct we'd need to separate it into two checks:
ENABLE_ASAN and ENABLE_UBSAN, but the convenience of combining them into
the same flag outweighs the theoretical correctness.
Also note in CONTRIBUTING.md that debug builds in ASAN is not supported.
Technically it is the debug runtime that is not supported, which cmake
automatically enables when using the debug build type. However, neovim
can't be built with debug builds without linking to the debug runtime
since the third party libraries has likely been linked to the debug
runtime if it was built with debug build type. This technicality is
likely uninteresting to the potential developer and it's easier to just
say to use a release build type.
(cherry picked from commit ef311fcd12)
Problem: `functional/vimscript/api_functions_spec` skips a test if the
runtime files are not generated, but this check was broken in a
refactor.
Solution: Since runtime files are now generated for all test targets, do
not skip this test.
(cherry picked from commit 2df8a81e21)
This is a maintenance release, focusing on bugfixes.
Notably, fixes were made for using and testing Nvim on less
common platforms, like big endian systems.
BUG FIXES
- defaults: Set 'fsync'
- Dont create data dir if it's a broken symlink
- Make InspectTree handle nested injection
- Remove nested for_each_tree in TSTreeView
- api: Use a conditional stack for nvim_cmd
- change: Update fold after on_bytes
- deps: Bump libmpack from luarocks
- extmark_to_array: Convert conceal_char to bytes
- highlight: Apply 'winblend' to float border
- inccommand: Save and restore '[ and '] marks
- inccommand: Don't crash with "split" and 'n' flag
- lsp: Handle NUL bytes in popup text
- lua: Correct return value for on_key with no arguments
- lua: Crash in nlua_error
- man: Make :Man with a range work
- mappings: Fix mapset() not replacing map with backslash
- marks: Handle switching buffer properly
- mouse: Avoid dragging after click label popupmenu callback
- mouse: Avoid dragging when clicking next to popupmenu
- move: Check the correct buffer
- options: Setting 'scroll' with resized grid
- quickfix: Check if delete failed in qf_fill_buffer()
- set_bool_option: Use int value to make bool Object
- set_option_from_tv: Use v_bool to access boolean option
- terminal: Assign channel to terminal earlier
- terminal: Avoid Insert mode in Terminal buffer
- terminal: Don't lose focus on <MouseMove>
- terminal: Keep focus when scrolling number column of another window
- terminal: Make backslashes in 'shell' work on Windows
- treesitter: Foldexpr tweaks
- treesitter: Remove incompatible c query (again)
- ui: Empty line before the next message after :silent command
BUILD SYSTEM!
- Fix universal mac builds
- ci: Ensure correct headers are used on macOS
DOCUMENTATION
- List NVIM_APPNAME as env on manpage
- starting.txt: Correct step number
PERFORMANCE
- Remove redundant strlen in skipwhite
TESTING
- Fix dependencies between test cases and test flakiness
The releases doesn't work on intel mac as libintl isn't available on the system
by default. This makes `:language` not work for the shipped macos releases,
though the reduction in build system complexity most likely outweighs that.
Currently, the release build picks up headers in
`/Library/Frameworks/Mono.framework/Headers`. You can verify this by
downloading the latest nightly build and checking the output of `nvim
--version`.
These headers are likely to be from a different version of `libintl` than the
one we link to. Let's avoid usage of them by setting `CMAKE_FIND_FRAMEWORK` to
`NEVER`.
All releases that aren't directly maintained by us should live in the
`neovim/neovim-releases` repository to make it clear that neovim isn't
directly responsible for it and to correctly manage expectations that
it's provided on a best-effort case.
Closes https://github.com/neovim/neovim/issues/26717.
Allow a "*count" suffix in a screen line to repeat the screen line for
"count" times.
The change is made to Screen:expect() and Screen:get_snapshot() instead
of Screen:render() so that screen expectations generated using code can
still work and test failures can still be readable.
A snapshot is now also printed on failure so that there is no need to
run the test again with Screen:snapshot_util().
Problem: reg_executing() returns wrong result in :normal with range
when 'showcmd' is set (after 8.2.4705).
Solution: Reset "pending_end_reg_executing" when executing a register.
closes: vim/vim#13707615202bd0e
(cherry picked from commit 2d1143f285)
These are meant for expensive or situational tests that may not be
suitable to run each time, but can occasionally be useful.
Currently only add testing for s390x. The job is enabled by adding the
ci-s390x label in github.
(cherry picked from commit 20dd9f3a26)
Cmake 3.28+ will fail if two projects download the same file to prevent
scheduling problems. This can be circumvented by downloading luajit to a
unique location for each target. This is theoretically non-optimal since
we need to download the same file three times for universal builds, but
universal builds are rare and the convenience of this method outweighs
setting up the dependencies optimally. This fixes the currently broken
release workflow for mac.
Closes https://github.com/neovim/neovim/issues/26526.
(cherry picked from commit d5c656c7d0)
Problem: screenpos() may crash with neg. column
Solution: validate and correct column
closes: vim/vim#13669ec54af4e26
(cherry picked from commit 6052fac625)
runtime(syntax): unlet b:filetype_in_cpp_family for cpp & squirrel
Update runtime/syntax/cpp.vim and runtime/syntax/squirrel.vim to unlet
b:filetype_in_cpp_family as it remains set even after updating the ft of
a file manually or through a modeline, not allowing c specific keywords
to be highlighted.
Since the variable b:filetype_in_cpp_family is only used by the c.vim
syntax script, unlet it directly after sourcing the c.vim runtime file
instead of at the end of the script.
Also update the last Change Header for both files.
closes: vim/vim#13650ff0baca865
Co-authored-by: laburnumT <laburnumtec@gmail.com>
(cherry picked from commit 6333be96ca)
fix(log): increase size of buffer for nvim instance name
16 bytes is not enough room for existing usage of the buffer by the
tests, so the name may get truncated and cause log_spec test to fail:
FAILED test/functional/core/log_spec.lua @ 30: log messages are formatted with name or test id
test/helpers.lua:146: retry() attempts: 51
test/helpers.lua:155: Pattern "%.%d+%.%d/c +server_init:%d+: test log message" not found in log (last 100 lines): Xtest_logging:
ERR 2023-11-24T23:36:34.252 T1274.2445945.0 server_init:57: test log message
ERR 2023-11-24T23:36:34.275 T1274.2445945.0 server_init:57: test log message
(cherry picked from commit ec80e4cb4d)
Co-authored-by: James McCoy <jamessan@jamessan.com>
Problem: 'breakindent' is not drawn after diff filler lines.
Solution: Correct check for whether 'breakindent' should be drawn.
closes: vim/vim#13624588f20dece
Problem:
With vim.treesitter.foldexpr, `o`-ing two lines above a folded region
opens the fold. This does not happen with legacy foldexprs. For example,
make a markdown file with the following text (without indentation),
enable treesitter fold, and follow the instruction in the text.
put cursor on this line and type zoo<Esc>
initially folded, revealed by zo
# then this fold will be opened
initially folded, revealed by o<Esc>
Analysis:
* `o` updates folds first (done in `changed_lines`), evaluating
foldexpr, and then invokes `on_bytes` (done in `extmark_splice`).
* Treesitter fold allocates the foldinfo for added lines (`add_range`)
on `on_bytes`.
* Therefore, when treesitter foldexpr is invoked while running `o`, it
sees outdated foldinfo.
Solution:
`extmark_splice`, and then `changed_lines`. This seems to be the
standard order in other places, e.g., `nvim_buf_set_lines`.
