Reorder comments, function attributes, and macro def args

[elcritch]

Some changes I've wanted for a while. Still a bit of a WIP.

• reorder comments to convert C-style / Doxygen prefixed comments to Nim-style postfixed comments
  • currently only tested on proc defs, and struct bodies
• add in __attribute__ to list of declarations for functions to help with parsing function attributes
• add --def arg to allow defining a macro from the command line
• add --stdints pre-defined mangle to convert C stdint to Nim ints

Example usage:

c2nim \
  --strict --header \
  --stdints # gives: --mangle:'{u?}int{\d+}_t=$1int$2' \
  --reordercomments \
  --def:RCL_PUBLIC='__attribute__ ()' \
  --def:RCL_WARN_UNUSED='__attribute__ ()' \
    testsuite/tests/rcl_arguments.h

Example re-order comments:

  /** * allocator objects.  */
  void * state;

becomes:

  void * state; ##\
      ## allocator objects. 

Parsing __attributes__ and --def: makes it posible to handle the GCC C function pragmas style. For example --def:RCL_WARN_UNUSED='__attribute__ ()' and --def:RCL_PUBLIC='__attribute__ ()' will make the below parse:

// function defined with macros attributes:
RCL_PUBLIC RCL_WARN_UNUSED int rcl_arguments_get_count_unparsed(const rcl_arguments_t * args);
// expands to:
__attribute__(public) __attribute__(...) int rcl_arguments_get_count_unparsed(const rcl_arguments_t * args);

That last change is responsible for a good portion of C files not parsing properly. Eventually it'd be nice to add pragmas to the Nim procs so people could define them, but that's a project for another day.

[elcritch]

Ok, I think this is ready. The new stuff has tests. Also added:

• --render:extranewlines adds in extra newlines before proc defs. This plumbs
  in renderFlags to the CLI parser

[arnetheduck]

does this really need to be an option? ie why not always do this (since it's needed to make the comments compatible with nim doc)

[elcritch]

Two reasons: 1) to make it easier to get accepted 2) because I’d have to reorder the comments in all the tests.

Though perhaps it could default to on and use the option to disable it. Plus there may be cases it’s wrong. I sorta hack the new lines with a heuristic.

[arnetheduck]

this would feel more natural as a --no-stdints - ie the quality of the wrapper is significantly improved when these types are recognised and it's hard to imagine code that uses these names for anything else than the standard interpretation