Summary:ASTERISK-13949: [patch] no audio: chan_mobile with windows mobile 6.1 needs HEADSET_AGW, not HANDSFREE_AGW
Reporter:spblinux (spblinux)Labels:
Date Opened:2009-04-14 05:26:57Date Closed:2009-04-21 13:25:15
Versions:Frequency of
Environment:Attachments:( 0) 013-search_sdp-chan_mobile.c.patch
( 1) search_sdp-chan_mobile.c.diff
Description:connecting windows mobile 6.1 phone to chan_mobile (tested with xperia x1 and with SimValley XP-25 alias Rover PC C6 / P6):
- mobile search in asterisk cli to get rfcomm port
- writing this port into mobile.conf
- receiving calls works but audio goes through mobile (phone does not open sco connection)
- replacing the port given by 'mobile search' (=HANDSFREE_AGW_PROFILE_ID)
  with the port found for HEADSET_AGW_PROFILE_ID makes audio go through asterisk
- incoming and outgoing calls work with 'headset audio gateway' rfcomm port

(In case of xp-25 headset_agw is rfcomm port 1, handsfree_agw rfcomm port is 2)


concerns refactored 1.6.2.x version and older versions of chan_mobile

patch will be attached to add headset_agw to 'mobile search'
Comments:By: Matthew Nicholson (mnicholson) 2009-04-14 10:57:50

Did you test with 'type=headset'?  Using HEADSET_AGW_PROFILE_ID should not work properly.  We really do want HANDSFREE_AGW_PROFILE_ID or HEADSET_PROFILE_ID.

By: spblinux (spblinux) 2009-04-14 15:47:17

I did not use type headset. - Normal setup for the two windows mobile 6.1 phones (x1, xp-25 mentioned above) results in audio routed through phone speaker and phone microphone. Everything else works normally.

(I do use other phones with chan_mobile which work with normal setup - and stop working with HEADSET_AGW_PROFILE_ID...). It seems to be a windows mobile 6.1 issue.

Side note to refactored version: the use of an asterisk smoother in mbl_write works fine on embedded device
(mipsel router); however the rfcomm parsing logic seems to have timing issues because the refactored
chan_mobile segfaults on my (slow 200MHz cpu) router. - As I do not have the time to produce and study
coredumps I leave this as a side note and do not open a separate bug.