Commit Graph

5507 Commits

Author SHA1 Message Date
Julian Orth
60a7ad8a47 remove HAVE_SETJMP_H 2014-03-31 21:24:59 -03:00
oni-link
f9b5ca8b85 Remove feature HAVE_QSORT
qsort conforms to C99, so we don't need our own version.
2014-03-31 21:24:59 -03:00
Nicolas Hillegeer
7b14caf604 alloc: xmalloc and xrealloc can't return null
With this annotation/function attribute, we can inform gcc that these
functions never return null, which can save a lot of redundant null checks.
2014-03-31 18:42:17 -03:00
Nicolas Hillegeer
22e17e25f1 func-attr: add some extra handy function attrs
The attributes in question are:

- nonnull: specify whether a function argument cannot/may not be null
- returns_nonnull: specify whether a function will not return a null
  pointer (example: xmalloc can't return null, so it should be annotated as
  such). Only available from gcc 4.9 onwards.

Currently these attributes are only supported by gcc.
2014-03-31 18:42:17 -03:00
Nicolas Hillegeer
f5db0d01ba func-attr: clang 3.5 doesn't support alloc_size
The attribute was removed in commit c047507 in the clang repository as it
was never properly implemented anyway. This fixes compiling with clang 3.5.

Fixes issue #429
2014-03-31 18:42:17 -03:00
John Schmidt
5d30654512 Rename ga_init2() to ga_init() 2014-03-31 14:42:11 -03:00
John Schmidt
e02790cad4 Inline ga_init() 2014-03-31 14:42:11 -03:00
Thiago de Arruda
a4c77ab5f0 Fix shell.{c,h} comments to match the style guide 2014-03-31 09:20:24 -03:00
Thiago de Arruda
c7d3e85dc8 Add doxygen comments to the functions in shell.c 2014-03-31 08:34:27 -03:00
Thiago de Arruda
ab61c2caa7 Rewrite show_shell_mess as a ShellOpts flag 2014-03-31 08:07:47 -03:00
Thiago de Arruda
206a38871f Update call_shell signature to use ShellOpts 2014-03-31 08:07:47 -03:00
Thiago de Arruda
0d61b1c470 Refactor SHELL_* defines into enum typedef
The SHELL_* defines are the bitflags that can be passed to `mch_call_shell`.
The enum is defined in 'os/shell.h', where all shell-related functions will
eventually be defined.
2014-03-31 08:07:47 -03:00
Thiago de Arruda
607e1c7ee4 Refactor/cleanup argument parsing functions 2014-03-31 08:07:47 -03:00
Thiago de Arruda
f496d619a9 Extract shell_build_argv from mch_call_shell 2014-03-31 08:07:47 -03:00
Thiago de Arruda
c791922224 Make extra_shell_arg a mch_call_shell parameter 2014-03-31 08:07:46 -03:00
Thiago de Arruda
8a61c27b1e Extract shell_count_argc from mch_call_shell 2014-03-31 08:07:46 -03:00
Thiago de Arruda
2d28251a6e Extract shell_skip_word from mch_call_shell 2014-03-31 08:07:46 -03:00
Thiago de Arruda
1ab6cf47bd Split code for counting and parsing arguments 2014-03-31 08:07:46 -03:00
Thomas Wienecke
54782ecfe0 Add doxygen-style comments to functions from os/fs.c. 2014-03-31 07:36:00 -03:00
Felipe Oliveira Carvalho
0e998066b2 xrealloc(): similar to xmalloc()
Replaced all calls to realloc by xrealloc. All `== NULL` tests can be removed
and the code within `!= NULL` tests can be unwrapped.
2014-03-31 07:31:47 -03:00
Felipe Oliveira Carvalho
7bdd1f1898 Document xmalloc() and deprecate lalloc() 2014-03-31 07:31:47 -03:00
Felipe Oliveira Carvalho
5afc1161ca Avoid a seg fault when exiting after OOM error
I'm debugging OOM behavior using http://www.nongnu.org/failmalloc/ on Linux.

gdb environment:

```
set env LD_PRELOAD=libfailmalloc.so
set env FAILMALLOC_SPACE=0xF00000
```

SEGV was happening like this:

```
Starting program: /home/felipe/code/neovim/build/bin/nvim
[Thread debugging using libthread_db enabled]
Using host libthread_db library "/lib/x86_64-linux-gnu/libthread_db.so.1".
Vim: Error: Out of memory.

Program received signal SIGSEGV, Segmentation fault.
0x00000000004d3719 in getout (exitval=1) at
/home/felipe/code/neovim/src/main.c:836
836       if (*p_viminfo != NUL)
(gdb)
```

After the fix it works as expected:

```
(gdb) set environment LD_PRELOAD=libfailmalloc.so
(gdb) set environment FAILMALLOC_SPACE=0xF00000
(gdb) r
Starting program: /home/felipe/code/neovim/build/bin/nvim
[Thread debugging using libthread_db enabled]
Using host libthread_db library "/lib/x86_64-linux-gnu/libthread_db.so.1".
Vim: Error: Out of memory.
Vim: Finished.
[Inferior 1 (process 21505) exited with code 01]
(gdb)
```
2014-03-31 07:31:47 -03:00
Felipe Oliveira Carvalho
de1575f3ea xmalloc() that succeeds or gracefully aborts 2014-03-31 07:31:47 -03:00
Justin M. Keyes
d9f5cd6290 Re-integrate FEAT_SIGNS code, close #383
- omit FEAT_NETBEANS_INTG and FEAT_SIGN_ICONS
- omit FEAT_GUI blocks
2014-03-29 20:47:39 -03:00
Nicolas Hillegeer
28b3659955 use function attributes on allocation functions
This should allow gcc/clang/icc to generate better code and clarifies
function intent.
2014-03-29 20:10:06 +01:00
Nicolas Hillegeer
d5708b200a add func_attr.h to be able to use func attribs
Currently enables these for every gnu-like compilers, i.e.: gcc, clang and
icc.
2014-03-29 20:09:57 +01:00
John Schmidt
44e40b02cf Move remove_duplicates to garray.c 2014-03-29 13:29:53 -03:00
John Schmidt
f5154d7451 Extract path.c from misc1.c 2014-03-29 13:29:52 -03:00
Thomas Wienecke
4e29a820b6 Address clint warnings and other style issues.
* Add const.
* Fix conditions (move && from end to start of line).
* Use int32_t instead of long.
* Use //-style comments.
2014-03-27 19:57:55 -03:00
Thomas Wienecke
5762c4e528 Rename mch_* functions to os_* in os module. 2014-03-27 19:57:55 -03:00
Thiago de Arruda
3f7011ab91 Fix prepare watcher teardown in the event loop 2014-03-27 19:22:44 -03:00
Thiago de Arruda
e995b21567 Re-integrate FEAT_FILTERPIPE code
This feature was accidentally removed when doing the initial import from vim. It
makes vim use pipes instead of temporary files for filtering buffers through
shell commands.

I found that this was missing when looking for references of
SHELL_READ/SHELL_WRITE outside mch_call_shell`.

When `mch_call_shell` is reimplemented on top of libuv process management
facilities, pipes will always be used for communication with child processes so
it makes sense to enable the feature permanently.
2014-03-27 17:36:32 -03:00
Thiago de Arruda
1e8eb4e2c6 Make EventType enum private to input.c
This enum doesn't need to be public since `event_poll` is only interested in
user input(but other events may be handled by libuv callbacks).
2014-03-27 17:25:31 -03:00
John
7326894a06 Cleanup of indent.c to make it comply with style guide.
- Uncrustify indent.c
- Fix comments
- Fix whitespace
- Format get_lisp_indent (needs to be refactored completely)
- Fix endline comments
- Change to postincrement
- Fix clint.py errors
- Move static function last
- Fix blank endlines
2014-03-27 11:55:56 -03:00
Thiago de Arruda
452804638d Reimplement input functions on top of libuv
The functions `mch_inchar`, `mch_breakcheck`, `mch_char_avail` were
reimplemented on top of libuv. Here's how it works:

- When Neovim needs to wait for characters, it will transfer control to libuv
  event loop.
- When the libuv event loop gets user input, it will transfer control back to
  Neovim
- Neovim uses the `input_read` function to get the actual data read by libuv.

With this scheme its possible to keep Neovim single-threaded while enjoying the
benefits provided by libuv.

This commit leaves SIGWINCH broken for now
2014-03-26 18:30:18 -03:00
Felipe Oliveira Carvalho
f8432ef127 Use realloc instead of vim_realloc 2014-03-26 18:29:17 -03:00
Felipe Oliveira Carvalho
580ababcb5 HAVE_AVAIL_MEM is always undefined. Remove tests and dead code
Defining HAVE_AVAIL_MEM doesn't even build. The code tries to call
`mch_avail_mem` which is not defined.
2014-03-26 18:29:17 -03:00
Felipe Oliveira Carvalho
83161200c4 Remove MEM_PROFILE related code
Code around `#ifdef MEM_PROFILE` was used to profile vim's memory
comsumption. It's very likely broken as new code is using malloc() and free()
directly.

In this day and age, valgrind can solve in a much reliable way what
this code was trying to do.
2014-03-26 18:29:17 -03:00
Felipe Oliveira Carvalho
bd78282940 Use memcmp instead of vim_memcmp 2014-03-26 16:28:13 -03:00
Felipe Oliveira Carvalho
07dad7acf3 Use memmove instead of mch_memmove 2014-03-26 16:28:13 -03:00
Petter Wahlman
dbc904956a Prefer enums over incremental defines 2014-03-26 16:26:56 -03:00
Felipe Oliveira Carvalho
ed3e9d51ac Use memset instead of vim_memset
Ran some commands on terminal, reviewed changes, and made some manual changes
too.

    find src | xargs perl -i -p -e "s/vim_memset/memset/g"
    git grep -l memset | egrep "\.c$" | xargs perl -i -p -e \
        's/(#include "vim\.h")/#include <string.h>\n\n\1/g'
2014-03-25 18:57:02 -03:00
Felipe Oliveira Carvalho
68bc6bce29 Do not include farsi.c in main.c
- Change CMakeLists.txt to compile farsi.c normally
 - Add const to global variables in farsi.h and define them in farsi.c (no need
   to include farsi.h with DO_INIT defined in globals.h)
 - Include farsi.h where necessary
 - Include all necessary headers in farsi.c
 - Move farsi function declarations from main.h to farsi.h
2014-03-25 18:48:45 -03:00
Felipe Oliveira Carvalho
b7f953b9e4 Move some #defines from farsi.h to farsi.c 2014-03-25 18:48:45 -03:00
Thiago de Arruda
32f118a47f Implement mch_delay on top of libuv
Needed to temporarily move two static variables from os_unix.c to 'globals.h'
as those are shared by other functions still in os_unix.
2014-03-24 19:25:07 -03:00
Felipe Oliveira Carvalho
ed42c808b6 inline arabic_char() 2014-03-24 18:23:24 -03:00
Felipe Oliveira Carvalho
5c66dfea3c Reformat new function in arabic.c to comply with style guide 2014-03-24 18:23:24 -03:00
Felipe Oliveira Carvalho
8b2a5e678e Move arabic.h #defines from arabic.h to arabic.c
- Move arabic_shape() decl from main.h to arabic.h
 - Move arabic_combine() and arabic_maycombine() from mbyte.c to
   arabic.c as these functions use the #defines I moved.
 - Remove the unnecessary include of arabic.h in globals.h
 - Remove include of arabic.c (sic) in main.c (change CMakeLists.txt to compile
   arabic.c normally)
2014-03-24 18:23:24 -03:00
Felipe Oliveira Carvalho
fd6fd35eef Turn ARABIC_CHAR into a function 2014-03-24 18:23:23 -03:00
John Szakmeister
7bd4d68d93 Don't require busted.
Only provide the unittest target if busted was found.  And only build
nvim-test if the unittest target exists by excluding nvim-test from all.
Note: this means nvim-test won't be built by default, but it will be
built when you try to run unittests.
2014-03-23 05:42:18 -04:00