reverting broad cosmetic/style change because:
- increases merge-conflicts
- increases overhead of merging upstream Vim patches
- reasons for change are ambiguous, so default to no change
By simpler cases I mean cases where the OOM error is not expected to be handled
by the caller of the function that calls `alloc`, `lalloc`, `xrealloc`,
`xmalloc`, `alloc_clear`, and `lalloc_clear`.
These are the functions that:
- Do not return an allocated buffer
- Have OOM as the only error condition
I took note of the functions that expect the caller to handle the OOM error and
will go through them to check all the callers that may be handling OOM error in
future commits.
I'm ignoring eval.c and ex_.c in this series of commits. eval.c will soon be
obsolete and I will deal with ex_.c in later PRs.
The last occurrence of `RealWaitForChar` was replaced by the `os_microdelay`
function. `mch_new_shellsize` had an empty body, so there seems to be no reason
for keeping it around
This removes all signal handling code from os_unix.c to os/signal.c. Now signal
handling is done like this:
- Watchers for signals are registered with libuv default event loop
- `event_poll` continuously calls `poll_uv_loop` to produce events until it
receives user input, SIGINT or a timeout
- Any signals received in `poll_uv_loop` will push events to a queue that is
drained and processed by `event_poll`
Signals aren't handled directly in the libuv callback to avoid recursion in the
event loop(which isn't supported by libuv).
The same principle will apply to other events in the future: Push to a queue
from a libuv callback and drain it from `event_poll`
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
See #137 for the issue.
Every header in the proto directory was:
* Given include guards in the form
#ifndef NEOVIM_FILENAME_H
#define NEOVIM_FILENAME_H
...
#endif /* NEOVIM_FILENAM_H */
* Renamed from *.pro -> *.h
* Moved from src/proto/ to src/
This would have caused conficts with some existing headers in src/;
rather than merge these conflicts now (which is a whole other can of
worms involving multiple and conditional inclusion), any header in src/
with a conflicting name was renamed from *.h -> *_defs.h (which may or
may not actually describe its purpose, the change is purely a
namespacing issue).
Once all of these changes were made a script was developed to determine
what #includes needed to be added to each source file to describe its
dependencies and allow it to compile; because the script is so short
and I'll just list it here:
#! /bin/bash
cd $(dirname $0)
# Scrapes `make` output for provided error messages and outputs #includes
# needed to resolve them.
# $1 : part of the clang error message between filename and identifier
list_missing_includes() {
for file_missing_pair in $(CC=clang make 2>&1 >/dev/null | sed -n "s/\/\(.*\.[hc]\).*$1.*'\(.*\)'.*/\1:\2/p"); do
fields=(${file_missing_pair//:/ })
source_file=${fields[0]}
missing_func=${fields[1]}
# Try to find the declaration of the missing function.
echo $(basename $source_file) \
\#include \"$(grep -r "\b$missing_func __ARGS" | sed -n "s/.*\/\(.*\)\:.*/\1/p")\"
# Remove duplicates
done | sort | uniq
}
echo "Finding missing function prototypes..."
list_missing_includes "implicit declaration of function"
echo "Finding missing identifier declarations..."
list_missing_includes "use of undeclared identifier"
Each list of required headers was added by hand in the following format:
#include "vim.h"
#include "*_defs.h"
#include "filename.h"
/* All other includes in same module here, in alphabetical order. */
/* All includes from other modules (e.g. "os/*.h") here in alphabetical
* order. */
cproto (http://invisible-island.net/cproto/) was used to do the bulk of
the work in batch; even the most recent version had some issues with
typedef'd parameters; a quick "patch" was to modify `lex.l` to
explicitly include all vim typedefs as known types. One example from
`vim.h` is
typedef unsigned char char_u;
which was added in `lex.l` as
<INITIAL>char_u { save_text_offset(); return T_CHAR; }
Even with these changes there were some problems:
* Two files (`mbyte.c` and `os_unix.c`) were not cleanly converted.
* Any function with the `UNUSED` macro in its parameter list was not converted.
Rather than spend more time fixing the automated approach, the two files
`mbyte.c` and `os_unix.c` were converted by hand.
The `UNUSED` macros were compiler specific, and the alternative, generic
version would require a different syntax, so in order to simplify the
conversion all uses of `UNUSED` were stripped, and then the sources were
run back through cproto. It is planned to reconsider each use of
`UNUSED` manually using a new macro definition.
- Cleanup source tree, leaving only files necessary for compilation/testing
- Process files through unifdef to remove tons of FEAT_* macros
- Process files through uncrustify to normalize source code formatting.
- Port the build system to cmake
