RE: [Iaxclient-devel] iaxclient lib (mainly libiax2) changes..

[Steven S. <ss...@so...>]

Steve,=20

I downloaded the tarball with (I assume) the code changes.  I am still
both loosing live calls and failing to receive incoming calls.  I am
running on Windows XP with the source compiled in GCC under MingGW.
This only appears to be an issue under IAX2.

Thanks,

Steve

-----Original Message-----
From: iax...@li...
[mailto:iax...@li...] On Behalf Of Steve
Kann
Sent: Monday, December 29, 2003 1:50 PM
To: iax...@li...
Cc: Babar Shafiq; Steve Underwood
Subject: [Iaxclient-devel] iaxclient lib (mainly libiax2) changes..


I've applied a few patches to the library:

1) First, I applied the two patches that SteveU sent to me, as they were
delivered.=A0 Thanks so much, Steve.=A0 I'm thinking it's highly likely =
this
will fix the problems that people have been reporting on the asterisk
mailing list (but not here too much), where:
=A0=A0 a) asterisk is timing out live calls, because of ping timeouts =
(the
pings were a bit late, and were getting=20
=A0=A0=A0 =A0=A0=A0 dropped).
=A0=A0 b) The client was missing incoming calls at times.


2) Second, I applied changes to address the Win32 build issues outlined
in the patch that Babar submitted on 19 December (Thanks, Babar!). The
changes I applied weren't exactly as he had them in the patch, but I
think that I addressed each of the issues, with one exception.=A0 Going
through them:

a) sendto type mismatch:=A0 I made an alternate declaration of sendto_t
which I believe should match the Win32 header declaration better.=A0 =
[this
is a proper fix, instead of casting to int].

b) His change to the order of elements in struct { } buf inside of
iax.c:=A0 I think that this change is _incorrect_.=A0 Can someone who =
knows
MSVC help out with this?=A0 I did not apply this.=A0 Here's the relevant
section from his diff:
616,619d615
<         struct {
<                 struct iax_frame fr2;
<                 unsigned char buffer[4096];             /* Buffer --
must preceed fr2 */
<         } buf;
620a617,620
 =20
>         struct {
>                 unsigned char buffer[4096];             /* Buffer --
must preceed fr2 */
> 		  struct iax_frame fr2;
>         } buf;
   =20
Basically, this is all Mark's doing :)=A0 What mark is trying to do is
allocate an extra 4K past the end of fr2, so that we can use the
zero-length afdata member at the end of the frame with 4k of room.=A0 =
MSVC
complains here at the declaration stage.


c) __FUNCTION__ not defined in MSVC:=A0 Only used __FUNCTION__ is =
__GNUC__
is defined.=A0 (maybe I could have just done ifdef __FUNCTION__ there?

d) #define snprintf to _snprintf..

e) structure packing stuff.=A0 Did basically the same thing, but without
totally repeating all the declarations.=A0 Just surrounded them with
appropriate preprocessor-fu to make both MSVC and GNUC happy.

f) inline define.

g) add time.h to iaxclient_lib.h, instead of just to winfuncs.h

I've only tested this by compiling here on Linux so far.=A0 If someone
could be so kind as to try this out on Win32 and especially with MSVC.

-SteveK