| Age | Commit message (Collapse) | Author |
|---|
|
The standard says "if an encoding error occurs ... the conversion state
is unspecified".
|
|
|
|
|
|
Close #647
|
|
|
|
|
|
See https://reasonml.github.io/docs/en/comparison-to-ocaml.html
for major differences between Reason and OCaml.
|
|
|
|
|
|
|
|
Fix #543
Close #588
|
|
|
|
|
|
Previously something like n<C-e> would also apply the count `n` to the
subsequent action.
|
|
|
|
With POSIX ERE the pattern ^$ matches strings ending with a new
line because an empty match is reported after the trailing newline
at the very end of the input.
This is undesirable for use cases like
x g/^$/ d
which is supposed to delete all empty lines of a file.
As a fix we disregard empty matches at the end of the given range.
|
|
|
|
|
|
As seen in #669 I didn't *get it* one needs to have a special stub in
`visrc.lua` for everything to work properly. Putting at least a few
words in the man page might help some others like me in the future :)
Close #671
|
|
|
|
In ncurses 6.1, the TERMINAL structure was updated[0] to store data in `int`
instead of `short`, and terminfo definitions for 256-color terminals were
updated from `pairs#32767` to `pairs#0x10000`.
However, since vis stores the value of COLOR_PAIRS in a short (ncurses
internally stores it as an int), it is now overflowing into negative, breaking
color support completely.
The standard `init_pair` entry points still use `short` for their parameters, so
just restrict the pairs to `SHRT_MAX` during allocation.
[0] http://invisible-island.net/ncurses/announce-6.1.html#h4-new-library
|
|
|
|
* fix character literals (#\" no longer quotes the entire file etc.)
* properly nest block comments and support simplified datum comment
* add r7rs keywords, functions and directives
* fix identifiers
- pipes were not recognized as delimiters
- some valid identifiers were not recognized
- some were partially parsed as keywords
- quoting only worked on plain alphanumeric identifiers
* fix numbers (some valid numbers were not recognised)
* dont parse boolean constants as functions
|
|
|
|
text_object_word() was only sending the last part of a pathname to
vis-complete.
text_object_longword() is better, but sometimes may send a bit too much,
so leading delimiters for some languages are stripped additionally.
|
|
The previous commit would generate duplicate entries when files with the
same name exist in more than one directory.
|
|
|
|
Move the "case $PATTERN" block inside "if $COMPLETE_WORD" to make it
clear that it is specific only to the "else" branch.
Eliminate the $START variable - it was used only once, and using
dirname(1) directly is obvious enough.
Remove the comments inside the "case" block (explaining the "what") and
replaced them with a single comment explaining the "why".
|
|
|
|
This keeps the existing testing infrastructure, which pipes newline
separated commands to stdin, working.
|
|
Fix #656
|
|
|
|
Fix #607
|
|
Close #655
|
|
|
|
This allows to end PHP code sections. Otherwise token is treated as part of the comment and parser continues to parse whatever is after.
|
|
The color settings are currently stored in the `vis.lexers` table,
make sure it is not nil even when loading the lexer module (or one
of its dependencies e.g. lpeg) failed.
|
|
|
|
|
|
At the command prompt, commands were not recognized properly when the
cursor was placed on a delimiting character (:, ?, /) while pressing
enter.
Fix #653
|
|
When the shell cannot find any matching files, the glob is not expanded,
and vis-open will return the absolute path of the current working
directory (because dirname outputs '.'), followed by the filename,
followed by a literal '*'. This commit checks that the final path
actually exists, and if not, exits with status 1.
It also uses text_object_longword for the range to match, so that
absolute paths are accepted, and replaced properly (else it only works
back to the first '/').
|
|
|
|
Fix #644
|
|
Correctly handle single letter words surrounded by special symbols, e.g. [c].
Fix #643
|
|
|
|
When the fore and background colors are the same, swapping them has
no effect. Instead use the specified cell attributes. Previously the
CELL_ATTR_REVERSE used in the default selection style was ignored.
In general the default style definitions for non-Lua builds could
probably be improved further.
Fix #635
|
|
This should allow the :-command prompt to be closed with <C-c><C-c>
(from insert mode) or <C-c> (from normal mode).
Fix #608
|
|
|
|
Fix #636
|
|
|
