[MPlayer-cvslog] r27496 - in trunk/stream: asf_mmst_streaming.casf_streaming.c http.c librtsp/rtsp_session.cnetwork.c network.h pnm.c rtp.c stream.cstream_ftp.c stream_netstream.c stream_rtsp.c tcp.c udp.c

[JonY]

Zuxy Meng wrote:
> "Dominik 'Rathann' Mierzejewski"<dominik at rangers.eu.org>  写入消息新闻:20080831082928.GA11794 at mokona.greysector.net...
>> On Saturday, 30 August 2008 at 18:21, Zuxy Meng wrote:
>>> "JonY"<10walls at gmail.com>  写入消息新闻:48B96416.4040206 at gmail.com...
>>>> diego wrote:
>>>>> Author: diego
>>>>> Date: Sat Aug 30 00:55:39 2008
>>>>> New Revision: 27496
>>>>>
>>>>> Log:
>>>>> Move duplicated '#define closesocket close' into network.h along with
>>>>> network-related #include #ifdeffery.
>>>>>
>>>>>
>>>>> Modified:
>>>>>      trunk/stream/asf_mmst_streaming.c
>>>>>      trunk/stream/asf_streaming.c
>>>>>      trunk/stream/http.c
>>>>>      trunk/stream/librtsp/rtsp_session.c
>>>>>      trunk/stream/network.c
>>>>>      trunk/stream/network.h
>>>>>      trunk/stream/pnm.c
>>>>>      trunk/stream/rtp.c
>>>>>      trunk/stream/stream.c
>>>>>      trunk/stream/stream_ftp.c
>>>>>      trunk/stream/stream_netstream.c
>>>>>      trunk/stream/stream_rtsp.c
>>>>>      trunk/stream/tcp.c
>>>>>      trunk/stream/udp.c
>>>>>
>>>> Hi,
>>>> It looks like one of the hunks broke MinGW build.
>>>>
>>>> libmpdemux and the codecs under it cannot be compiled properly because
>>>> definitions from "./loader/wine/*.h" is conflicting with the native
>>>> headers. Conflicts with "loader/wine/windef.h" is especially visible.
>>>>
>>>> "ad_internal.h", used by many of the files under libmpdemux includes
>>>> "stream/stream.h", which then includes "stream/network.h" if
>>>> "CONFIG_NETWORK" is defined, which then in turn includes the native
>>>> "ws2tcpip.h".
>>>>
>>>> r27495 builds with minor workarounds (read: local hacks) for
>>>> HAVE_WINSOCK2_H and closesockets.
>>>>
>>>> Any ideas to fix the problem?
>>> I'd say we'd better revert this mess first.
>> Why don't we have a look at the exact compilation failures first?
>> The original commit was a major cleanup and I'd hate to see it
>> reverted if it can be easily fixed.
>
>
> There're just too many. To name a few:
>
> 1. winsock2.h complains about lacking of definition for LPGUID, because
> codec-cfg.h only defines GUID.

Include ws2tcpip.h first? Both have the same length, different member 
names though.

 From winnt.h
typedef struct _GUID {
	unsigned long  Data1;
	unsigned short Data2;
	unsigned short Data3;
	unsigned char  Data4[8];
} GUID, *REFGUID, *LPGUID;

 From codec-cfg.h
typedef struct {
	unsigned long f1;
	unsigned short f2;
	unsigned short f3;
	unsigned char f4[8];
} GUID;

> 2. Definitions of INT32 and boolean conflicts with jpeglib.
> 3. STREAM_SEEK conflicts with stream/stream.h
Rename STREAM_SEEK?

> 4. Lots of redefinition where any header under loader/wine/ is included.
>
> We should include<winsock2.h>  only when it's needed, because it in turn
> includes<windows.h>  which is *big* mess.
>

IMHO, this should be fixed so native MinGW interfaces are preferred over 
WINE interface when building under MinGW.

I'm curious why wine headers are used at all for MinGW builds.