Russell Bryant [Mon, 14 May 2007 19:13:00 +0000 (19:13 +0000)]
Properly handle AST_CONTROL_PROGRESS by just ignoring it. An unknown indication
will trigger an error and cause sounds to stop, which in this case, is ringing.
Olle Johansson [Mon, 14 May 2007 18:52:09 +0000 (18:52 +0000)]
Handle network errors, like host or network unreachable, in a better way. This means that
calls to hosts or qualify (OPTION) messages will fail quicker if the TCP/IP stack tells us
that there is an issue.
Since this is an unconnected UDP socket, we will not get error messages directly
in most cases, but maybe on the second and third try.
Only perform stripping of - strings from the channel name for Zap channels. Anywhere else we might remove a legitimate part of a device name. (issue #9668 reported by stevedavies)
Russell Bryant [Fri, 11 May 2007 16:05:43 +0000 (16:05 +0000)]
When MD5 authentication is not possible because there is no challenge present,
either because the Challenge action was never issued, or some other reason,
give a proper error message and return an error instead of claiming that the
user wasn't found.
(reported by jsmith on IRC)
Russell Bryant [Thu, 10 May 2007 22:23:42 +0000 (22:23 +0000)]
Strip terminal escape sequences from CLI command output that is going to be
sent out over the manager interface.
(issue #9659, reported by pari, fixed by me)
Russell Bryant [Wed, 9 May 2007 16:55:27 +0000 (16:55 +0000)]
Modify ast_senddigit_begin() to use the same assumptions used elsewhere in the
code in that if a channel does not have a send_digit_begin() callback, it only
cares about DTMF END events. (pointed out by Michael Neuhauser on the
asterisk-dev list)
Properly handle hints that point to multiple devices in chan_sip. Why chan_sip is even doing this I have no idea but I would rather not go into a rant. (issue #9536 reported by rlister)
Russell Bryant [Wed, 9 May 2007 16:43:50 +0000 (16:43 +0000)]
Only call ast_senddigit_begin() in ast_senddigit() if the channel has a
send_digit_begin() callback. Checking the END_DTMF_ONLY flag was the
wrong thing to do, because that flag indicates that a *bridged* channel
only wants DTMF END events coming from this channel.
Russell Bryant [Wed, 9 May 2007 13:24:03 +0000 (13:24 +0000)]
I have seen multiple people post questions trying to figure out what the
message "The configure script must be executed before running 'make'" means.
So, add another like that says to specifically run ./configure. If this isn't
obvious enough, then they should be using something like AsteriskNOW and not
installing from source.
Merged revisions 62945,63402,63519 via svnmerge from
https://origsvn.digium.com/svn/asterisk/branches/1.2
........
r62945 | crichter | 2007-05-03 17:39:21 +0200 (Do, 03 Mai 2007) | 1 line
when we're in state WAITING4DIGS, we use the asterisk tone-generator which prods us, so we can't just return -1 in misdn_write in this case. Added a MISDN_KEYPAD channel variable, and fixed the sending of keypad. this enables us to modify the call forward parameters in the switch.
........
r63402 | crichter | 2007-05-08 17:07:37 +0200 (Di, 08 Mai 2007) | 1 line
added application misdn_check_l2l1 which tries to pull up the L1/L2 on all ports that have the layers down in a group. It waits then for a timeout. This helps for scenarios where multiple PMP BRIs are grouped together, or where a provider has a faulty PTP Implementation, that looses the L2 after a while.
........
r63519 | crichter | 2007-05-09 13:26:16 +0200 (Mi, 09 Mai 2007) | 1 line
release_chan frees ch, so we should never touch ch after release_chan, this may cause segfaults.
........
When a peer is seeded or built tell the devicestate core to update it's status. This is easier then having chan_sip load before pbx_config. (issue #9658 reported by dlynes)
improve loader a bit, by avoiding trying to initialize embedded modules twice and avoiding trying to load modules from disk when they have been loaded already during the 'preload' pass (reported by blitzrage on IRC, patch by me)
Russell Bryant [Thu, 3 May 2007 15:23:13 +0000 (15:23 +0000)]
Fix YADB (Yet Another DTMF Bug) ((C) Russell Bryant, 2007, TM, Patent Pending).
This set of changes came from a debugging session I had with Dwayne Hubbard.
When he called into his home FXO, ran the Echo application, and pressed a
digit, the digit would be echoed back and would never end. This is fixed,
along with a couple other little improvements.
* When chan_zap is in the middle of playing a digit to a channel, it feeds
back null frames, not voice frames. So, I have modified ast_read to check
the timing on emulated DTMF when it receives null frames, in addition to
where it was doing this on voice frames.
* Make a tweak to setting the duration on emulated DTMF digits. If there was
no duration specified, it set it to be the minimum, instead of the default.
* Instead of timing the emulated digits off of the number of samples in audio
frames that pass through, just use time values. Now there is no code in this
section that assumes 8kHz audio.
some fixes for PMP Hold/Retrieve, it should work now, when briding=no
........
r61770 | crichter | 2007-04-24 15:50:05 +0200 (Di, 24 Apr 2007) | 1 line
added lock for sending messages to avoid double sending. shuffled some empty_chans after the cb_event calls, this avoids that a release_complete from a quite different call releases a fresh created setup by accident.
........
r62885 | crichter | 2007-05-03 15:59:00 +0200 (Do, 03 Mai 2007) | 1 line
fixed the problem that misdn_write did not return -1 when called with 0 samples in a frame this resultet in a deadlock in some circumstances, when the call ended because of a busy extension. added encoding of keypad.
........
increase reliability and efficiency of static Realtime config loading via ODBC:
don't request fields we aren't going to use
don't request sorting on fields that are pointless to sort on
explicitly request the fields we want, because we can't expect the database to always return them in the order they were created
(reported by blitzrage in person (!), patch by me)
improve static Realtime config loading from PostgreSQL:
don't request sorting on fields that are pointless to sort on
use ast_build_string() instead of snprintf()
don't request the list of fieldnames that resulted from the query when we both knew what they were before we ran the query _AND_ we aren't going to do anything with them anyway
(patch by me, inspired by blitzrage's bug report about res_config_odbc)
Russell Bryant [Wed, 2 May 2007 22:59:09 +0000 (22:59 +0000)]
Merge changes from team/russell/inband_dtmf ...
Fix some issues related to generating inband DTMF. There are two changes here:
1) The list of DTMF tones in the senddigit_begin() function explicitly
specified 100ms of the tone followed by 100ms of silence. This really
broke things with the way that Asterisk now wants complete control
over when the digit begins and ends. So, regardless of what Asterisk
really wanted to do, this was going to play out the tone at the length it
wanted to. This caused various problems like DTMF translation to inband to
be extremely unreliable.
The list of tones has been changed so that the correct DTMF tone is played
indefinitely until Asterisk tells it to stop.
2) ast_write() had to be modified to let a DTMF_END frame get processed even
when a generator is present. This is how the tone will finally get stopped.
(issues #8944, #9250, #9348, maybe others. Thanks to mdu113 from #8944 for
the testing and feedback!)
Issue 9638 - if a text frame is sent with no terminating NULL through a bridged
IAX connection, the remote end will receive garbage characters tacked onto the
end.
Steve Murphy [Wed, 2 May 2007 17:10:50 +0000 (17:10 +0000)]
a)In chan_zap, set the clid, src fields in channel_alloc call. b)in the channel_alloc func, set the cid_num and name fields from the arglist[blush]. c) don't update the channel app & app data fields if you are in the 'h' extension. d)the load_module func in cdr_radius needs to return DECLINE, SUCCESS.
Russell Bryant [Tue, 1 May 2007 21:57:10 +0000 (21:57 +0000)]
Merged revisions 62547 via svnmerge from
https://origsvn.digium.com/svn/asterisk/branches/1.2
........
r62547 | russell | 2007-05-01 16:55:19 -0500 (Tue, 01 May 2007) | 4 lines
Remove an unnecessary check that makes it so if you hang up after doing an
attended transfer before the target extension answers the channel, the transfer
is not successful. (issue #9338, patch by svanlund)
This patch fixes an issue where depending on the cause code, when the network
sends a PRI disconnect, the call may not be properly hung up.
(issue #9588, reported and patched by softins)
Russell Bryant [Mon, 30 Apr 2007 15:25:31 +0000 (15:25 +0000)]
When serving dynamic content, include a Cache-Control header to instruct the
browsers to not store the resulting content.
(issue #9621, reported by Pari, patch by me)
Russell Bryant [Fri, 27 Apr 2007 21:10:51 +0000 (21:10 +0000)]
Fix a weird problem where when a caller talking to someone sitting behind an
agent channel sent a digit, the digit would be played to the agent for forever.
This is because chan_agent always returned -1 from its send_digit_begin and _end
callbacks. This non-zero return value indicates to the Asterisk core that it
would like an inband DTMF generator put on the channel. However, this is the
wrong thing to do. It should *always* return 0, instead. When the digit begin
and end functions are called on the proxied channel, the underlying channel
will indicate whether inband DTMF is needed or not, and the generator will be
put on that one, and not the Agent channel.
(issue #9615, #9616, reported by jiddings and BigJimmy, and fixed by me)
Russell Bryant [Fri, 27 Apr 2007 16:14:11 +0000 (16:14 +0000)]
If no variables were passed into pbx_substitute_variables_helper_full(), then
don't even bother creating a temporary bogus channel, since that is only for
allowing certain functions to operate on the variables as if they were on a
channel. Most importantly, this fixes a crash.
(issue #9613, reported by callguy, fixed by me)
Issue #7351 - SIP Cancel fails due to the wrong contact uri. Reported by PPYY, failed to fix by OEJ
final fix by wojtekka - THANKS!!!! THis was a hard one to catch.
Revert previous fix for when the IAX2 channel goes funky (that's the technical term). This is causing legit calls to be prematurely hung up. (issue #9600 reported by justdave)
Merge rewritten group counting support. No more storing data on the variable list of the channels. That was bad, mmmk? (issue #7497 reported by sabbathbh)
Don't crash if a manager connection provides a username that exists in
manager.conf but does not have a password, and also requests MD5
authentication. (ASA-2007-012)
Russell Bryant [Tue, 24 Apr 2007 19:00:06 +0000 (19:00 +0000)]
Improve DTMF handling in ast_read() even more in response to a discussion on
the asterisk-dev mailing list. I changed the enforced minimum length of a
digit from 100ms to 80ms. Furthermore, I made it now enforce a gap of 45ms in
between digits. These values are not configurable in a configuration file
right now, but they can be easily changed near the top of main/channel.c.
Russell Bryant [Tue, 24 Apr 2007 16:16:41 +0000 (16:16 +0000)]
Add a few more state changes in handle_frame_ownerless() so that the SLA code
will get notified of these changes even when an owner channel is not provided.
This isn't from a specific bug report, it's just something I noticed while
poking around.
Russell Bryant [Mon, 23 Apr 2007 18:17:00 +0000 (18:17 +0000)]
Some dialplan functions, such as CUT(), expect to operate on variables on a
channel. So, this little hack lets them work in places where a channel doesn't
exist, such as within DUNDi configuration.
(issue #9465, reported and patched by Corydon76, testing by blitzrage)
Russell Bryant [Mon, 23 Apr 2007 17:57:32 +0000 (17:57 +0000)]
Ensure that digits passing through Asterisk have a reasonable minimum length.
It is currently 100 ms. If someone thinks this should be different, feel free
to speak up. (related to issues #8944, #9250, and #9348)
If the '* to hangup' option is not enabled, we don't need to disable * as a valid exit key.
If it was enabled, this statement would've never been checked in the first place.
Russell Bryant [Fri, 20 Apr 2007 18:19:18 +0000 (18:19 +0000)]
Fix the UpdateConfig manager action to properly treat "variables" and "objects"
differently (a=b versus a=>b).
(issue #9568, reported by pari, patch by me)
Bug 9557 - Specifying the GetVar AMI action without a Channel parameter can
cause Asterisk to crash. The reason this needs to be fixed in the functions
instead of in AMI is because Channel can legitimately be NULL, such as when
retrieving global variables.
projects / thirdparty / asterisk.git / log
