That's a very good advice, except (personally though) maybe for
autotools.
One point about #define vs. enum - I used to like former, but now I
like the latter, there is a middle ground:
enum {
KEY_UP = 100
};
#define KEY_UP KEY_UP
This way, preprocessed .c/.cpp keeps the value, and can be reused
quickly for environments which can parse "C" syntax, to the point that
they can import any "C" declaration. (luajit ffi for example)
http://luajit.org/ext_ffi_tutorial.html
The latter one (#define) then is useful for other such systems that
automatically import values, or maybe they have only preprocessor,
rather than full blown "C" parser.
In any case, it was real easy job wrapping libzmq for luajit ffi (note:
there is more mature lua binding for libzmq already in another lib,
which is mentioned in the zeromq site):
https://github.com/malkia/ufo/blob/master/ffi/zmq.lua
A note: There is something fishy about WINDOWS ERROR CODES in the ZMQ
library.
I don't think checking for #ifndef E_SOMETHING, then #define
E_SOMETHING (ZMQ_HAUSNUMERO + MagicValue) is safe.
Certain E_xxx errors are defined in one Windows SDK, and not in WDKs
A better approach is to always declare them
Even better would be if errors are the same across all platforms (less
work for binding writers).
On 11/3/2011 7:09 AM, Martin Sustrik wrote:
> Nice summary:
>
> https://git.kernel.org/?p=linux/kernel/git/kay/libabc.git;a=blob_plain;f=README
>
> Martin
_______________________________________________
zeromq-dev mailing list
[email protected]
http://lists.zeromq.org/mailman/listinfo/zeromq-dev
