Updated usage.

darcs-hash:20070228172037-55b16-0f4fdb3d30f034f83a1625e66c00c9f620bdc338.gz

Port 0 used as ephemeral port when bind()ing.

darcs-hash:20070228171850-55b16-3b068983d4adb2d67832f5e723143bdd472e17ed.gz

The % constructions and $< do not work with pmake (used on BSDs).

darcs-hash:20070228170315-55b16-ac51e4b5293462fc4f9891068995d7067f0b5682.gz

darcs-hash:20070226202518-55b16-331cd75ce2ea1dd2c49b766a3a9b86b66cd2c904.gz

darcs-hash:20070226185621-55b16-9f23ba87cc902d6ca57db6eec26d6e4096d0c4e8.gz

darcs-hash:20070226185603-55b16-3dc06c411c2443868b1a9610efe632ed0b666ca4.gz

Sending un-SUBSCRIBE upon an error response to SUBSCRIBE, BYE to an error
response to PRACK.

darcs-hash:20070226185507-55b16-e86da5b3daf11e72aafe4630c3d3577316be5ae5.gz

darcs-hash:20070226185327-55b16-12ae40de1d406c0e33c570c61b7552e2f7b6f70d.gz

The termination request was sent occasionally sent twice, resulting in
assertion failure. Problem reported by Michael Jerris.

darcs-hash:20070226185130-55b16-bd97966b586a3a9a714e8426c2c017f3b4874184.gz

Now handle gracefully the case where the final response is delayed because
the PRACK to initial 100rel response has not been received and the client
CANCELs the INVITE request.

darcs-hash:20070226161956-55b16-d833253ff0d6bcd4c6df4a2812b70b8139bd5d0f.gz

darcs-hash:20070226141230-55b16-0c6ab00c73ec71e6c17e2bcb31831cf81c382f67.gz

darcs-hash:20070226141131-55b16-98eaa75242673eb47c6ac5a840176f5737bd8eee.gz

Problem reported by Michel Jerris.

darcs-hash:20070226141118-55b16-cb78dd9d6e63b1c6a07f131803b27f28ec889a4f.gz

Note that testing with 100rel/UPDATE cases is very incomplete.

darcs-hash:20070226140956-55b16-429cada445bdb59fdab4f90b5d3449a2c0b81cb7.gz

darcs-hash:20070223121343-55b16-9c5b45351506978cc3fcb2ac3434cfb85cd93ab6.gz

darcs-hash:20070223121205-55b16-d4ee29fef80472b0c413b1002a0e01e0b2005bf4.gz

darcs-hash:20070223111130-55b16-960db880673285f723ba03a897bc7d0a43735dee.gz

Stack crashed if an another request was sent using same dialog as the
initial request rejected by stack. Reported by Michael Jerris.

darcs-hash:20070223111055-55b16-87a055f868fb3c27a1f28aeb7146d3a4f41985f1.gz

Added su_root_name() and su_port_name() functions to API.

darcs-hash:20070222210558-88462-e6574b8fce46199f5c03887bad1ee66e9d2fcf9b.gz

Added extern int xyzzy;

darcs-hash:20070222210345-55b16-d83589c06a3aece82b4f1c81dd027c9bca068901.gz

darcs-hash:20070222210309-55b16-b521ea0b79b376263535c948c50f0561c491c25b.gz

darcs-hash:20070222205419-88462-0ffc6a66952d83fd366bef46ee30b51a05c19957.gz

darcs-hash:20070222205350-88462-5f0087b6fc994213be8caa0a98bf1827c107e27d.gz

darcs-hash:20070222203510-88462-a2606fb7baad6e7b2bba4dfe25a7d10e6887f2f9.gz

This caused an assertion failure in tport.c on Solaris.

darcs-hash:20070222203347-88462-92bb0549818eb67da9b3de4cb192c78e898e16a4.gz

- Autoconf now checks for support of %zu and %llu
- Autoconf disables the E_BAD_PTR_INT_COMBINATION warning on SUNWspro
- Fixed inline support:
  - SU_INLINE and SU_HAVE_INLINE macros defined correctly if compiler does
    not support inline
  - Using su_inline in public headers
  - Added normally inlined functions to library
  - Added some inlined functions as macros
- Using explicit signedness with bit fields
  - (The bitfields in soa_session.h should go away, but...)
- Fixed warnings in nta.c
- Defining missing SOL_SCTP in tport_type_sctp.c

Michael Jerris sent patch for some of the changes mentioned above.

darcs-hash:20070222203018-88462-ec89b84a178f699bcd1e2195cfa75e2bbe10c9ae.gz

darcs-hash:20070222151957-7659e-20bfc00e1936bbe1b0afb28700162f4d92bf1c59.gz

darcs-hash:20070218233718-1b897-719890eea47b86c238c02f1a62f6d0cc6c6f4eee.gz

darcs-hash:20070218233516-1b897-6c3711af1f41b7b596f0749f69a3a9691fca94ba.gz

darcs-hash:20070218233339-1b897-ddb02216d8007dfefd7e282b44257af180ee0980.gz

darcs-hash:20070218233217-1b897-a97ef32e08b8dcd59467fca9140c063cf71a10a5.gz

darcs-hash:20070218233156-1b897-f198f079ec59f062f89fed7021301ed6491b0443.gz

darcs-hash:20070218233109-1b897-93604326f171ece6f14655b67a246332a08a4d7e.gz

According to the RFC 3261 state diagram the INVITE transaction is destroyed
immediately when a final 2XX series response is sent. Now nta returns a 481
response to CANCEL if it finds an INVITE server transaction that was
responded with a 2XX success response.

darcs-hash:20070214202456-55b16-26e78d85b04fa47ecf612966bd8766188390035a.gz

- dropping call in inconvenient times
- INVITE glare
- queueing re-INVITEs

darcs-hash:20070214202150-55b16-8466bad2111638f1e8bfb8c50dfce6d65a4903b2.gz

These problems are mostly introduced in 1.12.5:
- assert failed if INVITE was received when calling
  (reported by Michael Jerris)
- assert failed if nua_invite() was called second time before receiving
  final response to first INVITE (ditto)
- ACK is now always sent if session was terminated after receiving 2XX
- if nua has a pending incoming INVITE request, return 491 to
  nua_invite()

darcs-hash:20070214201847-55b16-962b2de246eee8100e4dba8d7a28c507b13c376c.gz

darcs-hash:20070214194256-55b16-b3204a32ab024950ee36e8f4d93d90a90095d246.gz

darcs-hash:20070212192239-2e7b9-6942b9de48eb88471c622dc01954015a47663fd1.gz

darcs-hash:20070212231038-7659e-af3d8e7848e33807a37ce091c2bc3d512b6c6801.gz

nua: Ignore CANCEL of incoming request if we have already sent a final response. Problem reported by Mike Jerris.

darcs-hash:20070212230200-7659e-b45cc5b2242a4aa9195c5ee1443784ad43d9708f.gz