[FFmpeg-devel] [PATCH] IPv6 support v.2

[Luca Abeni]

Hi Ronald

Ronald S. Bultje wrote:
> Hi,
> 
> On Nov 13, 2007 11:38 AM, Luca Abeni <lucabe72 at email.it> wrote:
> 
>>>> I generated the patches in a short time, so maybe some of them can have
>>>> typos or can be written in a better way, but I believe they show how to
>>>> remove duplicated code without (hopefully :) breaking anything.
>>>>
>>>> - address-len.diff removes some stupid ifdef by using the dest_addr_len
>>>>   field in both the IPv4-only and the generic code. In this way, we
>>>> decrease
>>>>   the differences between the two cases
> [..]
> 
> So now that all those are applied, what's step2?

My understanding was that you was going to post some updated patches after
I committed my ones (you said: "I'll say it this way: there's multiple
ways to get there, and I think your patches are a good start in one such
direction to fix the issue, and I'll gladly pick up from there and resolve
the whole thing").
But at this point I suggest you to wait for the libavnet changes before
refreshing your patches.


> I mean, all this
> re-arranging of a few lines is fun, but the rest of ffmpeg still doesn't
> speak ipv6, my patches are not going in

Maybe I just missed them (and in this case, let me apologize in advance),
but I did not see the updated patches after my changes.


> so that's apparently not right
> either, and there's no clear outline of where this is going. We need a plan
> and action. This "let's do nothing and wait for the Right Solution[tm] to
> come in" isn't working.
> 
> - in os_support.c, resolve_host() needs to be converted to be
> ipv6-compatible or removed

The "protocol independent code" should not use resolve_host(), I think...
If ffmpeg is not compiled with --disable-ipv6, udp.c does not use resolve_host().
tcp.c should probably be fixed to do the same.
I think there was an agreement on directly using getaddrinfo()?


> - all duplicate code in udp.c needs to be removed, right now it's just
> duplicated even more

Point out some duplicated code, and I'll try to do something about it.


> - tcp.c, ffserver, sdp.c, rtsp.c need to speak ipv6

The sdp.c part should be easy. Just send a patch for sdp.c (using getaddrinfo)
and I'll commit it.


> - api needs to be able to return sensible error codes and applications need
> to be able to report those

Once again, I am confused about this part: how is it related to IPv6?
Reporting sensible error codes should be something completely orthogonal to
the IPv6 stuff...


				Luca