[FFmpeg-devel] [PATCH] OS/2:Support linking against libcx

[KO Myung-Hun]

Hi/2.

Dave Yeo wrote:
> On 06/13/20 07:03 AM, KO Myung-Hun wrote:
>> Hi/2.
>>
>> Dave Yeo wrote:
>>> On 06/11/20 10:26 AM, Michael Niedermayer wrote:
>>>> On Wed, Jun 10, 2020 at 09:24:51PM -0700, Dave Yeo wrote:
>>>>> On 06/10/20 02:09 PM, Michael Niedermayer wrote:
>>>>>> On Tue, Jun 09, 2020 at 11:11:48PM -0700, Dave Yeo wrote:
>>>>>>> Hi, could I get this pushed to trunk and the 4.3 branch? Fixes a
>>>>>>> build break
>>>>>>> in libavformat/ip.c (implicit declaration of function
>>>>>>> 'getaddrinfo') and
>>>>>>> also need the prototype.
>>>>>>> Thanks,
>>>>>>> Dave
>>>>>> it seems this breaks build on linux
>>>>>
>>>>> Sorry about that, I'll test on Linux in the future.
>>>>> Here's a better patch as it doesn't touch configure.
>>>>> Thanks,
>>>>> Dave
>>>>
>>>> I can confirm this does not break build anymore, but iam not OS/2
>>>> maintainer nor do i have such box so ill leave review / application
>>>> to someone better suited for this
>>>>
>>>> thx
>>>> [...]
>>>
>>> Fair enough, I'll CC KOMH
>>
>> I have no problems at all with gcc 9.1.0 because FFmpeg already has
>> replacements for missing functions such as getaddrinfo().
>>
>> What is your build environment ? Maybe is libcx linked by default ?
>>
> 
> The problem only occurs if I link in libcx, LIBS=-lcx.
> Note that with libcx, the build system finds its getaddrinfo() and later
> fails due to no prototype and will also fail as part of the addrinfo
> struct uses a different type (char vs int IIRC). Without LIBS=-lcx, the
> build falls back to FFmpeg's builtin getaddrinfo().
> One of the main reasons to link to libcx is for the exceptq support.
> Also as it is recommended, others are likely to do the same thing.

Ok, then it would be be better to elaborate commit message.

> From f9fbdaaf6cdb6f886cbdf31c1983e452567cd857 Mon Sep 17 00:00:00 2001
> From: Dave Yeo <daveryeo at telus.net>
> Date: Tue, 9 Jun 2020 22:51:53 -0700
> Subject: [PATCH] OS/2:Support linking against libcx
> 
> Libcx contains extensions to libc such as getaddrinfo(), mmap() and poll(). While recommended to link against, it is optional
> 
> Signed-off-by: Dave Yeo <daveryeo at telus.net>
> ---
>  configure                | 1 +
>  libavformat/os_support.h | 3 +++
>  2 files changed, 4 insertions(+)
> 
> diff --git a/configure b/configure
> index 8569a60bf8..24ad990b52 100755
> --- a/configure
> +++ b/configure
> @@ -5997,6 +5997,7 @@ if ! disabled network; then
>      check_func inet_aton $network_extralibs
>  
>      check_type netdb.h "struct addrinfo"
> +    check_type libcx/net.h "struct addrinfo"

Check libcx/net.h first like:

    if check_headers libcx/net.h ; then
        check_type libcx/net.h "struct addrinfo"
    fi

Otherwise, platforms without libcx/net.h are treated as not having
`struct addrinfo' even if they have actually. I think, this is the cause
of build failure on linux.

>      check_type netinet/in.h "struct group_source_req" -D_BSD_SOURCE
>      check_type netinet/in.h "struct ip_mreq_source" -D_BSD_SOURCE
>      check_type netinet/in.h "struct ipv6_mreq" -D_DARWIN_C_SOURCE
> diff --git a/libavformat/os_support.h b/libavformat/os_support.h
> index 5e6b32d2dc..1904fc8d5d 100644
> --- a/libavformat/os_support.h
> +++ b/libavformat/os_support.h
> @@ -56,6 +56,9 @@
>  #  define fstat(f,s) _fstati64((f), (s))
>  #endif /* defined(_WIN32) */
>  
> +#if defined (__OS2__) && defined (HAVE_GETADDRINFO)
> +#include <libcx/net.h>
> +#endif
>

HAVE_GETADDRINFO does not guarantee libcx/net.h. Use

  #if HAVE_LIBCX_NET_H

guard. If you prefer, it's ok to check __OS2__ definition additionally.

FYI, HAVE_GETADDRINFO is always defined to either 0 or 1. So checking it
with `defined' is always true.

For HAVE_LIBCX_NET_H, you should add `libcx_net_h' to HEADERS_LIST.

Thanks.

-- 
KO Myung-Hun

Using Mozilla SeaMonkey 2.7.2
Under OS/2 Warp 4 for Korean with FixPak #15
In VirtualBox v6.1.10 on Intel Core i7-3615QM 2.30GHz with 8GB RAM

Korean OS/2 User Community : http://www.os2.kr/