- As a general rule of thumb one shouldn't use environment variables
for setting configuration options for CMake. The reason for this is
we don't know when CMake will be executed and re-evaluate that variable.
- e.g. If we run cmake a first time with a var set, and then run make on
a second session (with no var) and cmake is called because a
dependency changed, the option would be disabled
- This commit removes the use of environment vars from
src/nvim/CMakeLists.txt entirely
- Removed SKIP_UNITTEST since it could only be used to remove a target
at configuration time (and the target was optional anyway)
- Turned SANITIZE into an option, clang-asan.sh now passes cmake
-DSANITIZE=ON
- Removed SKIP_EXEC since it was disabling a target at configuration time
(not being used)
Nvim now relies much less on setting terminal mode to cooked mode, remove most
calls to settmode, except for those that happen on startup or when suspending.
Eventually even those will be handled by the UI layer.
- Call compiler from CMake instead of lua script to generate a
preprocessor file - allows for better/early error detection if
the compiler fails
- Preprocessor files are saved along with the headers as .i files
- Accept preprocessor lines with trailing chars after # as is
the case in Clang/Windows
- The fourth argument to gendeclarations.lua is now the path to
the proprocessor output file
Switch from POSIX's write() to fwrite(stdout,...) and disable buffering
since vim buffers output explicitly and flushes when needed, like when a
key is typed.
Problem : Command `s/\n//` is being translated into a call to do_join
with a count of 1. But do_join asserts its precondition count
>= 2, which is causing the program to abort.
Note : This in fact revealed bigger problems: generated join command
line count, as well as reported substitutions/lines were
wrong in several cases, since patch 7.4.232.
See:
[patch] http://markmail.org/message/vo7ruair5raccawp
[issue] https://code.google.com/p/vim/issues/detail?id=287
Solution : - Don't generate join command for single-line-range case.
- Make generated join command include:
* lines in range + 1, when range doesn't end at last line.
* lines in range, when range ends at last line.
- Make reported substitutions/lines always be
number-of-lines-joined - 1.
vim-patch:7.4.473
Problem: Cursor movement is incorrect when there is a number
column/sign/fold column and 'sbr' is displayed.
Solution: Adjust the column for 'sbr'. (Christian Brabandt)
https://code.google.com/p/vim/source/detail?r=v7-4-473
vim-patch:7.4.425
Problem: When 'showbreak' is used "gj" may move to the wrong
position. (Nazri Ramliy)
Solution: Adjust virtcol when 'showbreak' is set. (Christian Brabandt)
https://code.google.com/p/vim/source/detail?r=v7-4-425
** CID 74786: Resource leak (RESOURCE_LEAK)
/src/nvim/eval.c: 10614 in f_jobsend()
/src/nvim/eval.c: 10616 in f_jobsend()
save_tv_as_string() should return NULL and input_len <= 0 for an empty
string or error. Callers should check that input != NULL instead of
input_len > 0 and assert(input == NULL) when the length must be checked.
Also:
- Remove NO_CONSOLE_INPUT/NO_CONSULE preprocessor conditionals
- Remove ctrl_c_interrupts variable, check for mapped_ctrl_c directly in
process_interrupts()
- Move ui_inchar profiling to input_poll which is where Nvim blocks for input.
- By default vim_feedkeys escaped all input for CSI/K_SPECIAL bytes
before using it. However since vim_replace_termcodes() also escapes
the input string chaining these functions together escapes input twice
- vim_feedkeys() now takes a third Boolean argument to enable/disable
escaping
- Breaks API compatibility
- get_system_output_as_rettv() was missing a refcount increment when
returning an empty list, i.e. when there was no output
- we now use rettv_list_aloc() instead of list_alloc()
- issue #1530
`try_malloc` was changed in 8bb2c2c074 to
avoid a warning when size is 0. Then, this improves some things on that:
- Use local vars instead of changing parameters.
- Homogenize style for other related functions.
Two new functions, `event_enable_deferred()`/`event_disable_deferred()` have to
be called by code that is capable of handling asynchronicity. User-dialog states
like "press ENTER to continue" or the swap file confirmation no longer will
generate K_EVENT.
