vnc and lwip

[Jonathan Larmour]

Gary Thomas wrote:
> On Fri, 2004-05-28 at 13:39, Andrew Lunn wrote:
> 
>>On Fri, May 28, 2004 at 08:16:52PM +0100, Jonathan Larmour wrote:
>>
>>>Andrew Lunn wrote:
>>>
>>>>On Fri, May 28, 2004 at 06:33:55PM +0300, Jani Monoses wrote:
>>>>
>>>>
>>>>>Since lwip has a network.h header for compatibility with the other stacks
>>>>>this can be applied.Will do so if nobody objects.
>>>
>> 
>>
>>>Personally I dislike <network.h> when the new POSIX standard has a 
>>>standardised layout for the BSD sockets API. i.e. <sys/socket.h> etc.
>>
>>network.h is not normal, but it does sort out the macro magic needed
>>to get the header files to compile correctly. I think thats why Gary
>>decided to invent it.

Macro magic is an obstacle to using portable code without jumping through 
non-obvious (even if documented) hoops.

> Exactly, but with the new [FreeBSD] stack this is not so important.

Putting -D__ECOS=1 in the global CFLAGS might be better anyway so it 
appears in application flags that copy the global ones.

> It does still provide some useful stuff, including prototypes for
> non-standard functions.

Fair enough, although that should live under <cyg/...>. Given the way we 
define include paths we already clash with any user code that have their 
own network.h that they want to include even by #include "network.h". We 
should keep the top level directory free of non-standard headers.

Anyway, I'm too lazy to do anything about the existing situation :-), but 
I'd dislike any standardisation on something non-standard.

Jifl
-- 
eCosCentric    http://www.eCosCentric.com/    The eCos and RedBoot experts
--["No sense being pessimistic, it wouldn't work anyway"]-- Opinions==mine