Richard Mudgett [Tue, 13 May 2014 00:25:17 +0000 (00:25 +0000)]
chan_dahdi/sig_pri: Prevent unnecessary PROGRESS events when overlap dialing is enabled.
When overlap dialing is enabled, the lack of inband audio available
information in the SETUP_ACKNOWLEDGE events causes an interoperability
problem with SIP. sig_pri doesn't know if there is dialtone present when
a SETUP_ACKNOWLEDGE is received so it assumes it is there and posts an
AST_CONTROL_PROGRESS frame. The SIP channel driver then sends out a 183
Session Progress and blocks the desired 180 Ringing message when the
ALERTING message comes in.
* Made the configure script detect if the installed version of libpri
supports the SETUP_ACKNOWLEDGE enhancements.
* Using the new API, made generate an AST_CONTROL_PROGRESS frame on an
incoming SETUP_ACKNOWLEDGE message when the message indicates inband audio
is present instead of assuming that dialtone is present.
* Using the new API, made SETUP_ACKNOWLEDGE send out an inband audio
available indication only if dialtone is expected. The change also makes
the fallback behaviour of sending the PROGRESS message better by sending
it only if dialtone is expected.
* Changed receiving a PROCEEDING message to not generate an
AST_CONTROL_PROGRESS frame if the progress indication ie indicates
non-end-to-end-ISDN. This helps interoperability with SIP.
* Changed sending a PROCEEDING message in response to an
AST_CONTROL_PROCEEDING frame to not indicate inband audio available. It
was silly to do so anyway because the channel driver doesn't know if
inband audio is even available. This helps interoperability with SIP.
This patch and a corresponding change in libpri work together to allow
Asterisk to control the inband audio available progress indication ie on
the SETUP_ACKNOWLEDGE message when dialtone is present.
Joshua Colp [Sun, 11 May 2014 02:05:26 +0000 (02:05 +0000)]
framehooks: Add callback for determining if a hook is consuming frames of a specific type.
In the past framehooks have had no capability to determine what frame types a hook
is actually interested in consuming. This has meant that code has had to assume they
want all frames, thus preventing native bridging.
This change adds a callback which allows a framehook to be queried for whether it
is consuming a frame of a specific type. The native RTP bridging module has also
been updated to take advantange of this, allowing native bridging to occur when
previously it would not.
Joshua Colp [Sat, 10 May 2014 18:45:42 +0000 (18:45 +0000)]
framehooks: Add callback for determining if a hook is consuming frames of a specific type.
In the past framehooks have had no capability to determine what frame types a hook
is actually interested in consuming. This has meant that code has had to assume they
want all frames, thus preventing native bridging.
This change adds a callback which allows a framehook to be queried for whether it
is consuming a frame of a specific type. The native RTP bridging module has also
been updated to take advantange of this, allowing native bridging to occur when
previously it would not.
Kinsey Moore [Fri, 9 May 2014 22:39:22 +0000 (22:39 +0000)]
Allow Asterisk to compile under GCC 4.10
This resolves a large number of compiler warnings from GCC 4.10 which
cause the build to fail under dev mode. The vast majority are
signed/unsigned mismatches in printf-style format strings.
........
Merged revisions 413586 from http://svn.asterisk.org/svn/asterisk/branches/1.8
........
Merged revisions 413587 from http://svn.asterisk.org/svn/asterisk/branches/11
Jonathan Rose [Fri, 9 May 2014 16:35:16 +0000 (16:35 +0000)]
app_chanspy: Fix a bug where Barge mode could fail
If the barge audiohook was attached prior to the spyee and its peer
actually being bridged, the audiohook would not be applied and the
connected peer would not be able to hear audio from the spy when the
spy is in barge mode.
(closes issue ASTERISK-23381)
Reported by: Robert Moss
Review: https://reviewboard.asterisk.org/r/3505/
........
Merged revisions 413551 from http://svn.asterisk.org/svn/asterisk/branches/11
Richard Mudgett [Wed, 7 May 2014 20:58:03 +0000 (20:58 +0000)]
app_confbridge: Fixed "CBAnn" channels not going away.
Fixed a ref leak in conf_handle_talker_cb() everytime the conference
bridge was found to report a channel's talker status change. The
resulting leak caused the "CBAnn" channels and the conference bridge to
never be destroyed.
Thanks to Richard Kenner on the asterisk-user's list for locating the
problem.
Mark Michelson [Tue, 6 May 2014 22:00:31 +0000 (22:00 +0000)]
Improve XML sanitization in NOTIFYs, especially for presence subtypes and messages.
Embedded carriage return line feed combinations may appear in presence subtypes
and messages since they may be derived from user input in an instant messenger
client. As such, they need to be properly escaped so that XML parsers do not
vomit when the messages are received.
Mark Michelson [Tue, 6 May 2014 17:46:44 +0000 (17:46 +0000)]
Check for an act on failures to update contacts during registration.
There was an underlying issue in a realtime backend where database updates
would fail. Since we were not checking for failure, we would end up in a
strange state where the old database entry was still present but Asterisk
thought that it had been updated. Now when an entry fails to update, we
print a warning and delete the old contact from sorcery so there is no
mismatch between foreground and backend state.
Patches:
res_pjsip_registrar.patch by John Hardin (License #6512)
Mark Michelson [Fri, 2 May 2014 20:35:16 +0000 (20:35 +0000)]
Correct variable traversal logic in res_config_odbc's update_odbc function.
Closes issue ASTERISK-23675
Reported by Leando Dardini
Patches:
asterisk-23675-odbc-linkedlist-traversal_12.diff uploaded by Michael L. Young (license #5026)
Mark Michelson [Fri, 2 May 2014 20:02:53 +0000 (20:02 +0000)]
Return the number of rows affected by a SQL insert, rather than an object ID.
The realtime API specifies that the store callback is supposed to return the number
of rows affected. res_config_pgsql was instead returning an Oid cast as an int, which
during any nominal execution would be cast to 0. Returning 0 when more than 0 rows were
inserted causes problems to the function's callers.
To give an idea of how strange code can be, this is the necessary code change to fix
a device state issue reported against chan_pjsip in Asterisk 12+. The issue was that
the registrar would attempt to insert contacts into the database. Because of the 0
return from res_config_pgsql, the registrar would think that the contact was not successfully
inserted, even though it actually was. As such, even though the contact was query-able
and it was possible to call the endpoint, Asterisk would "think" the endpoint was unregistered,
meaning it would report the device state as UNAVAILABLE instead of NOT_INUSE.
The necessary fix applies to all versions of Asterisk, so even though the bug reported
only applies to Asterisk 12+, the code correction is being inserted into 1.8+.
Closes issue ASTERISK-23707
Reported by Mark Michelson
........
Merged revisions 413224 from http://svn.asterisk.org/svn/asterisk/branches/1.8
........
Merged revisions 413225 from http://svn.asterisk.org/svn/asterisk/branches/11
Richard Mudgett [Fri, 2 May 2014 16:33:59 +0000 (16:33 +0000)]
res_pjsip_refer: Add Referred-By header on INVITE for blind transfers.
Per rfc3892, the Referred-By header in a REFER must be copied into the
referenced request (IE. The outgoing INVITE to the transfer target).
* Automatically put the Referred-By header in the outgoing INVITE message
if the SIPREFERREDBYHDR channel variable is defined with a value.
* Made chan_sip.c:get_refer_info() set SIPREFERREDBYHDR for inheritance so
chan_pjsip has a better chance to interoperate.
* Fixed refer_blind_callback() and refer_incoming_refer_request() to not
modify the data in the pointer returned by pjsip_msg_find_hdr_by_name().
It seems wrong to modify that data since the calling routine doesn't own
the buffer.
Mark Michelson [Thu, 1 May 2014 15:41:43 +0000 (15:41 +0000)]
Remove unnecessary repetition checks from res_pjsip_exten_state
The PBX core already takes care of ensuring that repeated state changes
are not communicated to exten state consumers. Because the check in res_pjsip_exten_state
was incomplete, it was causing valid presence state changes not to be sent out. For instance,
if the presence state did not change but the message or subtype did, then no presence-related
NOTIFY request would be sent out.
closes issue ASTERISK-23672
Reported by Mark Michelson
* Fixed early exit in sip_msg_send() not destroying the message iterator.
* Made ast_msg_var_iterator_next() and ast_msg_var_iterator_destroy()
tolerant of a NULL iter parameter in case ast_msg_var_iterator_init()
fails.
* Made ast_msg_var_iterator_destroy() clean up any current message data
ref.
* Made struct ast_msg_var_iterator, ast_msg_var_iterator_init(),
ast_msg_var_iterator_next(), ast_msg_var_unref_current(), and
ast_msg_var_iterator_destroy() use iter instead of i.
* Eliminated RAII_VAR usage in res_pjsip_messaging.c:vars_to_headers().
........
Merged revisions 413139 from http://svn.asterisk.org/svn/asterisk/branches/11
chan_pjsip: Fix deadlock when retrieving call-id of channel.
If a task was in-flight which required the channel or bridge lock
it was possible for the synchronous task retrieving the call-id
to deadlock as it holds those locks.
After discussing with Mark Michelson the synchronous task was
removed and the call-id accessed directly. This should be safe as
each object involved is guaranteed to exist and the call-id will
never change.
Kinsey Moore [Wed, 30 Apr 2014 13:06:47 +0000 (13:06 +0000)]
Websocket: Add session locking and delay close
This resolves a race condition where data could be written to a NULL
FILE pointer causing a crash as a websocket connection was in the
process of shutting down by adding locking to websocket session writes
and by deferring session teardown until session destruction.
(closes issue ASTERISK-23605)
Review: https://reviewboard.asterisk.org/r/3481/
Reported by: Matt Jordan
........
Merged revisions 413123 from http://svn.asterisk.org/svn/asterisk/branches/11
res_stasis: Add progress indications to operations which perform media.
This change fixes operations which did not account for the fact that they may
be executed on channels which have not been answered. These operations will
now indicate progress when invoked.
ASTERISK-23560 #close
ASTERISk-23560 #comment Reported by: Jan Svoboda
res_pjsip_sdp_rtp: Fix issue where sending a hold SDP twice could cause an unhold.
This change fixes a bug where if an SDP with media address and sendonly was
received twice the underlying call would go off hold, instead of remaining on hold.
This occured because the code did not properly take into account that the SDP
may contain both a valid media address and the sendonly attribute.
The code now examines the sendonly attribute and media address first, so if the
SDP is received again no change will occur.
George Joseph [Tue, 29 Apr 2014 15:09:11 +0000 (15:09 +0000)]
Add "destroy" implementation for spinlock.
The original commit for spinlock was missing "destroy" implementations.
Most of them are no-ops but phtread_spin and pthread_mutex do need their
locks destroyed.
chan_pjsip: Implement core ability to get Call-ID of a channel.
This changes implement the "get_pvt_uniqueid" which is used to return the
technology specific unique identifier. In the case of SIP this is the Call-ID
of the dialog.
Kinsey Moore [Mon, 28 Apr 2014 20:01:02 +0000 (20:01 +0000)]
Bridging: Don't lock NULL bridges
When bridge locking was added for bridge snapshot creation, some
locations where bridge locking was added were not guaranteed to
actually have a bridge and locking NULL AO2 objects tends to cause
segfaults. This ensures that NULL bridges aren't locked.
Matthew Jordan [Fri, 25 Apr 2014 17:48:19 +0000 (17:48 +0000)]
res_rtp_asterisk: Add support for DTLS handshake retransmissions
On congested networks, it is possible for the DTLS handshake messages to get
lost. This patch adds a timer to res_rtp_asterisk that will periodically
check to see if the handshake has succeeded. If not, it will retransmit the
DTLS handshake.
Kevin Harwell [Thu, 24 Apr 2014 14:37:11 +0000 (14:37 +0000)]
pjsip realtime: increase the size of some columns
The string lengths on certain columns created through alembic for PJSIP were
too short. For instance, columns containing URIs are currently set to 40
characters, but this can be too small and result in truncated values. Added
an alembic migration script that increases the size of these columns and a
few others to 255.
ASTERISK-23639 #close
Reported by: Mark Michelson
Review: https://reviewboard.asterisk.org/r/3475/
George Joseph [Wed, 23 Apr 2014 20:06:03 +0000 (20:06 +0000)]
This patch adds support for spinlocks in Asterisk.
There are cases in Asterisk where it might be desirable to lock
a short critical code section but not incur the context switch
and yield penalty of a mutex or rwlock. The primary spinlock
implementations execute exclusively in userspace and therefore
don't incur those penalties. Spinlocks are NOT meant to be a
general replacement for mutexes. They should be used only for
protecting short blocks of critical code such as simple compares
and assignments. Operations that may block, hold a lock, or
cause the thread to give up it's timeslice should NEVER be
attempted in a spinlock.
The first use case for spinlocks is in astobj2 - internal_ao2_ref.
Currently the manipulation of the reference counter is done with
an ast_atomic_fetchadd_int which works fine. When weak reference
containers are introduced however, there's an additional comparison
and assignment that'll need to be done while the lock is held.
A mutex would be way too expensive here, hence the spinlock.
Given that lock contention in this situation would be infrequent,
the overhead of the spinlock is only a few more machine instructions
than the current ast_atomic_fetchadd_int call.
Richard Mudgett [Wed, 23 Apr 2014 18:00:40 +0000 (18:00 +0000)]
http: Fix spurious ERROR message in responses with no content.
Backport -r411687 and fix the fix because content_length is the length of
out plus the length of the file controlled by fd.
When a response has an out content length of 0, fwrite would be called to
write a buffer with no data in it. This resulted in the following classic
error message:
res_stasis: Fix crash when handling a failed blind transfer message.
This changes fixes a crash that occurs when stasis determines if it
should send a message out to an application or not. The code
incorrectly assumed that a bridge snapshot would always be present
when in reality for failure cases it may not be.
Kinsey Moore [Mon, 21 Apr 2014 16:15:34 +0000 (16:15 +0000)]
HTTP: Add TCP_NODELAY to accepted connections
This adds the TCP_NODELAY option to accepted connections on the HTTP
server built into Asterisk. This option disables the Nagle algorithm
which controls queueing of outbound data and in some cases can cause
delays on receipt of response by the client due to how the Nagle
algorithm interacts with TCP delayed ACK. This option is already set on
all non-HTTP AMI connections and this change would cover standard HTTP
requests, manager HTTP connections, and ARI HTTP requests and
websockets in Asterisk 12+ along with any future use of the HTTP
server.
Jonathan Rose [Mon, 21 Apr 2014 16:05:11 +0000 (16:05 +0000)]
chan_sip: Add sendrpid trust options
In r411189, some behavior was changed which made sendrpid behavior
act in a more trusting manner by sending full user data for peers
set with private caller presence in P-Asserted-Identity headers.
Since this changed long time expected behaviors, we decided to pull
that patch when that was pointed out by the community. Instead, this
patch provides a trust_id_outbound setting which will expose the data
per RFC-3325 if set to 'yes' and simply not send the PAI/RPID headers
at all if set to 'no'. By default trust_id_outbound will be set to
'legacy' which will preserve the behavior prior to these patches.
Extra special thanks to Walter Doekes for providing advice and
feedback.
(closes issue AST-1301)
(closes issue ASTERISK-19465)
Reported by: Krzysztof Chmielewski
Kinsey Moore [Mon, 21 Apr 2014 14:46:06 +0000 (14:46 +0000)]
Confbridge: Add references for kick all option
After the ability to kick all attendees from a conference was added, a
rework removed the comment about that feature from the CLI
documentation. This adds that documentation and adds "all" to the
participant tab completion list for the confbridge kick command.
Fix wrong dialtone. The "modulation" should not be referenced for tone+tone as it refers to the on-off characteristic - this often resulted in a single tone rather than the multitone as in the UK.
........
Merged revisions 412712 from http://svn.asterisk.org/svn/asterisk/branches/11
Matthew Jordan [Sat, 19 Apr 2014 02:13:15 +0000 (02:13 +0000)]
main/asterisk: Fix startup sequence for realtime features
When ASTERISK-23265/ASTERISK-23320 was fixed, it inadvertently led to realtime
features breaking. This was due to features loading prior to realtime. This
patch fixes this by loading features after loading dynamic modules.
Matthew Jordan [Sat, 19 Apr 2014 01:09:04 +0000 (01:09 +0000)]
app_sms: Fix uninitialized values; hangup channel when REL is sent successfully
This patch fixes two issues in app_sms:
(1) Firstly, the 'flags' field on the stack in sms_exec() is uninitialised,
causing it to use the wrong protocol in some cases. This patch correctly
initializes the flags fields.
(2) Secondly, when disconnect supervision is not working or
inbanddisconnect=yes is set in chan_dahdi.conf, app_sms was failing to
terminate the call after it sent the REL(ease) message and the peer stopped
talking to it. This patch fixes the code to handle the 'bad stop bit'
message more gracefully in that case, and hang up the call.
Review: https://reviewboard.asterisk.org/r/1392/
ASTERISK-18331 #close
Reported by: David Woodhouse
patches:
asterisk-fix-sms.patch uploaded by David Woodhouse (License 5754)
........
Merged revisions 412655 from http://svn.asterisk.org/svn/asterisk/branches/1.8
........
Merged revisions 412656 from http://svn.asterisk.org/svn/asterisk/branches/11
Jonathan Rose [Fri, 18 Apr 2014 18:54:53 +0000 (18:54 +0000)]
ARI: Make bridges/{bridgeID}/play queue sound files
Previously multiple play actions against a bridge at one time would cause
the sounds to play simultaneously on the bridge. Now if a sound is already
playing, the play action will queue playback to occur after the completion
of other sounds currently on the queue.
(closes issue ASTERISK-22677)
Reported by: John Bigelow
Review: https://reviewboard.asterisk.org/r/3379/
Rusty Newton [Fri, 18 Apr 2014 17:16:14 +0000 (17:16 +0000)]
sounds: Fix Sounds Makefile and XML that didn't support new sound prompt sets
In sounds/Makefile
1 Adds and moves some lines necessary for the en_GB core set. I'm just following how the other sets are defined here.
2 removes the ES extra sounds related lines as we don't have ES extra sound sets.
In sounds/sounds.xml
3 Adds member definitons for EN_AU, EN_GB, IT for core sound sets, and EN_GB in extra sound sets
Richard Mudgett [Fri, 18 Apr 2014 16:38:20 +0000 (16:38 +0000)]
Originated calls: Fix several originate call problems.
* Restore the reason value set by pbx_outgoing_attempt() to use
AST_CONTROL_xxx values as all the consumers were expecting rather than
cause codes.
* Fixed the dial routines to set cause codes for more than just
ast_request() so pbx_outgoing_attempt() reason codes will function.
* Fix inconsistent locked_channel return status in pbx_outgoing_attempt().
The chanel may not have been locked or the channel may have been a stale
pointer.
* Fixed the OutgoingSpoolFailed channel to run dialplan whenever the
dialing fails for an originate exten and 1 < synchronous.
* Fix incorrect ast_cond_wait() usage in pbx_outgoing_attempt().
Indroduced by issue ASTERISK-22212 patch.
* Made struct pbx_outgoing use the ao2 lock instead of its own lock for
the cond wait mutex. No sense in having two locks associated with the
same struct when only one is needed.
Richard Mudgett [Fri, 18 Apr 2014 16:19:17 +0000 (16:19 +0000)]
app_dial and app_queue: Make lock the forwarding channel while taking the channel snapshot.
* Fixed ast_channel_publish_dial_forward() not locking the forwarded
channel when taking the channel snapshot.
* Fixed app_dial.c:do_forward() using the wrong channel to get the
original call forwarding string.
* Removed unnecessary locking when calling ast_channel_publish_dial() and
ast_channel_publish_dial_forward() in app_dial and app_queue. Holding
channel locks when calling ast_channel_publish_dial_forward() with a
forwarded channel could result in pausing the system while the stasis bus
completes processsing a forwarded channel subscription.
Kinsey Moore [Fri, 18 Apr 2014 14:21:34 +0000 (14:21 +0000)]
ARI: Add debug logging for events and responses
This adds DEBUG level logging for ARI websocket events and HTTP
responses similar to what is available for AMI. Logging for ARI HTTP
requests is already adequate for debugging purposes.
res_pjsip: Handle reloading when permanent contacts exist and qualify is configured.
This change fixes a problem where permanent contacts being qualified were not
being updated. This was caused by the permanent contacts getting a uuid and not a
known identifier, causing an inability to look them up when updating in the
qualify code. A bug also existed where the new configuration may not be available
immediately when updating qualifies.
(closes issue ASTERISK-23514)
Reported by: Richard Mudgett
Jonathan Rose [Thu, 17 Apr 2014 21:47:10 +0000 (21:47 +0000)]
ARI: Add tones playback resource
Adds a tones URI type to the playback resource. The tone can be specified by
name (from indications.conf) or by a tone pattern. In addition, tonezone can
be specified in the URI (by appending ;tonezone=<zone>). Tones must be
stopped manually in order for a stasis control to move on from playback of
the tone. Tones may be paused, resumed, restarted, and stopped. They may
not be rewound or fast forwarded (tones can't be controlled in a way that
lets you skip around from note to note and pausing and resuming will also
restart the tone from the beginning). Tests are currently in development
for this feature (https://reviewboard.asterisk.org/r/3428/).
(closes issue ASTERISK-23433)
Reported by: Matt Jordan
Review: https://reviewboard.asterisk.org/r/3427/
Matthew Jordan [Thu, 17 Apr 2014 20:24:41 +0000 (20:24 +0000)]
main/Makefile: Fix build failure on SmartOS/Illumos/SunOS
This patch fixes two issues when building on SmartOS:
- channels/chan_oss.c: it makes sure soundcard.h is found
- main/Makefile: only use "-Wl,--version-script" when GNU LD is used as the Sun
Linker doesn't support that. Similar checks are already used elswhere in the
Makefile
Review: https://reviewboard.asterisk.org/r/3426
ASTERISK-23576 #close
Reported by: Sebastian Wiedenroth
patches:
fix-sunos.diff uploaded by Sebastian Wiedenroth (License 6597)
........
Merged revisions 412468 from http://svn.asterisk.org/svn/asterisk/branches/11
Kevin Harwell [Thu, 17 Apr 2014 15:16:42 +0000 (15:16 +0000)]
res_pjsip_refer: Channel variable SIPREFERTOHDR not being set during blind transfer
The SIPREFERTOHDR channel variable is not being set on any channel when
performing a blind transfer using PJSIP. The 'refer->refer_to' was not
being set during a blind transfer. Updated so the 'refer_to' is set to
the target uri on a blind transfer.
(closes issue ASTERISK-23502)
Reported by: John Bigelow
Review: https://reviewboard.asterisk.org/r/3445/
Richard Mudgett [Tue, 15 Apr 2014 17:56:53 +0000 (17:56 +0000)]
Remove unused RAII_VAR() declarations.
* Remove unused RAII_VAR() declarations. The compiler cannot catch these
because the cleanup function "references" the unused variable. Some
actually allocated and released resources that were never used.
* Fixed some whitespace issues in stasis_bridges.c.
The failing assertion ensures that the final snapshot gets generated so
CDR records can get finalized. The only place where a channel staging
snapshot flag could be left set is in chan_sip.c:handle_request_bye().
The function could return before clearing the flag because the channel
could dissappear while the function had to have the channel unlocked.
* Fixed handle_request_bye() channel snapshot staging coverage area to not
have a return in the middle of it and be unable to clear the staging flag.
* Pushed the channel snapshot staging coverage area into
ast_rtp_instance_set_stats_vars() to ensure that the staging is not
interrutped.
* Made callers of ast_rtp_instance_set_stats_vars() not call it with any
channels or channel driver private locks held to eliminate the deadlock
potential. The callers must hold references to the passed in channel and
rtp objects.
* Eliminated sip_hangup() trying to get the bridge peer. It is futile at
this point because the channel could never be in a bridge.
Richard Mudgett [Tue, 15 Apr 2014 16:36:38 +0000 (16:36 +0000)]
chan_sip.c: Moved some sip_pvt unrefs after their last use.
* Moved sip_pvt unref in ast_hangup() and handle_request_do() to the end
of the function. The unref needs to happen after the last use of the
pointer.
........
Merged revisions 412348 from http://svn.asterisk.org/svn/asterisk/branches/11
chan_sip: Send real CallerID information with P-Assserted-Identity (RFC-3325)
Prior to this patch, the P-Asserted-Identity header would include anonymous
caller id information which seems to go against the point of the
P-Asserted-Identity header. Now the real caller ID information will be
included in this header. Also, no privacy header would be included.
This patch adds 'Privacy: id' to outgoing SIP messages that include the
P-Asserted-Identity header.
(closes issue AST-1301)
---
........
Merged revisions 412328 from http://svn.asterisk.org/svn/asterisk/branches/1.8
........
Merged revisions 412329 from http://svn.asterisk.org/svn/asterisk/branches/11
autoservice acquires a local reference to the logger callid of each channel
in a loop. This local reference was not released, causing the callid of
every channel in autoservice to leak. This change moves the callid unref
inside the loop.
Kinsey Moore [Fri, 11 Apr 2014 12:35:52 +0000 (12:35 +0000)]
bridging: Ensure locking during snapshot creation
While the vast majority of bridge snapshot creation is locked properly,
there are currently some instances that are not. This adds the missing
locking to ensure bridge state is not malleable during snapshot
creation.
(closes issue ASTERISK-22904)
Review: https://reviewboard.asterisk.org/r/3415/
Reported by: Matt Jordan
Matthew Jordan [Fri, 11 Apr 2014 02:48:50 +0000 (02:48 +0000)]
main/astobj2: Make REF_DEBUG a menuselect item; improve REF_DEBUG output
This patch does the following:
(1) It makes REF_DEBUG a meneselect item. Enabling REF_DEBUG now enables
REF_DEBUG globally throughout Asterisk.
(2) The ref debug log file is now created in the AST_LOG_DIR directory.
Every run will now blow away the previous run (as large ref files
sometimes caused issues). We now also no longer open/close the file
on each write, instead relying on fflush to make sure data gets written
to the file (in case the ao2 call being performed is about to cause a
crash)
(3) It goes with a comma delineated format for the ref debug file. This
makes parsing much easier. This also now includes the thread ID of the
thread that caused ref change.
(4) A new python script instead for refcounting has been added in the
contrib/scripts folder.
Matthew Jordan [Thu, 10 Apr 2014 21:27:25 +0000 (21:27 +0000)]
res_hep_pjsip: Use the channel name instead of the call ID when it is available
During discussions with Alexandr Dubovikov at Kamailio World, it became
apparent that while the SIP call ID is a useful identifier prior to an Asterisk
channel being created, it is far more preferable to use the channel name (or
some channel based identifier) when the channel is available. Homer is smart
enough to tie the various messages together. This patch opts to use the channel
name when it is available, falling back to the call ID otherwise.
Kevin Harwell [Thu, 10 Apr 2014 21:07:44 +0000 (21:07 +0000)]
res_pjsip_pubsub: Set the body generation result to 0 for a valid path
The result of the "ast_sip_pubsub_generate_body_content" was not
set/initialized. Consequently, the nominal path potentially returned
an invalid value, thus not sending mwi notifications.
Mark Michelson [Wed, 9 Apr 2014 20:32:36 +0000 (20:32 +0000)]
Add a Command header to the AMI Mixmonitor action.
This fixes a parsing error that occurred during the processing of
the AMI action. The error did not result in MixMonitor itself
misbehaving, but it could result in the AMI response not giving
correct information back.
The new header allows for one to specify a post-process command
to run when recording finishes. Previously, in order to do this,
the post-process command would have to be placed at the end of
the Options: header.
Patches: mixmonitor_command_2.patch by jhardin (License #6512)
Richard Mudgett [Tue, 8 Apr 2014 21:23:46 +0000 (21:23 +0000)]
Internal timing: Add notice that the -I and internal_timing option are no longer needed.
Add notice messages during execution that the -I command line option and
the astersik.conf internal_timing option are no longer needed. The
internal timing functionality is now always enabled if there is a timing
module loaded.
NOTE: Since the command line options and the asterisk.conf config file are
processed before the logging system is initialized, the messages are
output to stderr.
Change requested as a result of asterisk-dev list comments about the
commit for ASTERISK-22846 that removed the -I and internal_timing options.
res_pjsip: Ignore explicit transport configuration if a WebSocket transport is specified.
This change makes it so if a transport is configured on an endpoint that is a WebSocket
type the option will be ignored. In practice this is fine because the WebSocket
transport can not create outgoing connections, it can only reuse existing ones. By
ignoring the option the existing PJSIP logic for using the existing connection will
be invoked and stuff will proceed.
(closes issue ASTERISK-23584)
Reported by: Rusty Newton
Kinsey Moore [Mon, 7 Apr 2014 20:39:55 +0000 (20:39 +0000)]
PJSIP: Ensure test event has new state
The change that fixed the pubsub test event's use of a dangling pointer
also changed when it was processed relative to the pjsip subscription
state change processing. This change corrects the order of events while
holding a reference to the pointer that was previously dangling.
Jonathan Rose [Mon, 7 Apr 2014 16:02:44 +0000 (16:02 +0000)]
AGI/Manager: Prevent multiple NewExten events during AGI application changes
AGI applications would trigger NewExten events every time the state of the AGI
application changed. This has historically not been the behavior and this
behavior was introduced with a CDR patch. This patch corrects that.
(closes issue ASTERISK-23390)
Reported by: Benjamin Keith Ford
Review: https://reviewboard.asterisk.org/r/3406/
Kinsey Moore [Mon, 7 Apr 2014 14:28:41 +0000 (14:28 +0000)]
Stasis: Fix Stasis() bridge refcount issue
The Stasis() dialplan application monitors what bridge a channel is in
and so necessarily holds on to a bridge pointer. This change ensures
that it also holds on to a reference for that bridge to prevent the
bridge pointer from becoming a dangling pointer.
Kinsey Moore [Mon, 7 Apr 2014 13:24:09 +0000 (13:24 +0000)]
PJSIP: Fix crash introduced in r411671
The test event introduced in revision 411671 uses a dangling pointer to
access information about pubsub state changes. This moves the event to
within the lifetime of the pointer.
Richard Mudgett [Fri, 4 Apr 2014 19:02:57 +0000 (19:02 +0000)]
internal_timing: Remove the option and always make it enabled if a timing module is loaded.
The masquerade supertest frequently fails because either the local channel
chain doesn't completely optimize out or the DTMF handshake doesn't
completely get accross. Local channel optimization requires frames
flowing to trigger when optimization can happen. When optimization
happens the media frame that triggered the optimization is dropped.
Sending DTMF requires frames to flow in the other direction for timing
purposes while sending nothing. If internal timing is not enabled when
MOH is playing, Asterisk switches to received timing when an audio frame
is received. With optimization dropping media frames and MOH not sending
frames unless it receives frames, occasionaly there are no more frames
being passed and the test fails.
* The asterisk command line -I option and the asterisk.conf
internal_timing option are removed. Asterisk now always uses internal
timing when needed if any timing module is loaded. The issue
ASTERISK-14861 did this quite awhile ago in v1.4 but effectively is broken
if other internal timing modules besides DAHDI are used. The
ast_read_generator_actions() now only does received timing if it has no
choice for frame generators like MOH, silence, and playback streaming.
* Cleaned up some code dealing with frame generators in
ast_deactivate_generator(), generator_write_format_change(),
ast_activate_generator(), and ast_channel_stop_silence_generator().
Richard Mudgett [Fri, 4 Apr 2014 17:53:20 +0000 (17:53 +0000)]
Add some asserts that were handy when looking for a stasis cache problem.
* Assert if a channel is destroyed but has the snapshot staging flag set.
In this case the final channel destruction snapshot would never get taken.
* Assert if what we just got out of the stasis cache is not what we were
looking for. This assert would have saved several days searching for a
bug and a lot of my hair.
* Assert if the music on hold message posts could not find the associated
channel. A crash will happen later when manager tries to send the MOH AMI
message. This assert catches the problem when the stasis message is
posted instead of by the thread processing the defective message.
* Always generate a backtrace when an ast_assert() fails.
Matthew Jordan [Fri, 4 Apr 2014 15:11:48 +0000 (15:11 +0000)]
http: Fix spurious ERROR message in responses with no content
When a response has a content length of 0, fwrite would be called to write a
buffer with no data in it. This resulted in the following classic error
message:
Matthew Jordan [Thu, 3 Apr 2014 11:43:24 +0000 (11:43 +0000)]
res_hep: Fix crash when hep.conf not available
Parts of res_hep properly checked for a valid configuration object before
attempting to access the configuration. A check, however, was missed when
a packet is sent. This patch fixes the crash caused by not checking if the
configuration object is valid.
ASTERISK-23391 caused a regression where the symbol 'defaultlanguage'
was used by app_voicemail but not exported by main/asterisk. This
change renames the variable to ast_defaultlanguage. The variable was
already renamed in Asterisk 12+.
app_queue: Fix a bug where realtime members would be deleted during reload causing waiting callers to get ejected.
This patch causes realtime queue members to remain in queues during the reload process. Previously these
members would be removed causing any waiting callers to be ejected from the queue with a reason of "EXITEMPTY".
Matthew Jordan [Fri, 28 Mar 2014 18:09:03 +0000 (18:09 +0000)]
res_hep/res_hep_pjsip: Add a HEPv3 capture agent module and a logger for PJSIP
This patch adds the following:
(1) A new module, res_hep, which implements a generic packet capture agent for
the Homer Encapsulation Protocol (HEP) version 3. Note that this code is based
on a patch provided by Alexandr Dubovikov; I basically just wrapped it up,
added configuration via the configuration framework, and threw in a
taskprocessor.
(2) A new module, res_hep_pjsip, which forwards all SIP message traffic that
passes through the res_pjsip stack over to res_hep for encapsulation and
transmission to a HEPv3 capture server.
Much thanks to Alexandr for his Asterisk patch for this code and for a *lot*
of patience waiting for me to port it to 12/trunk. Due to some dithering on
my part, this has taken the better part of a year to port forward (I still
blame CDRs for the delay).
Alexandr Anikin [Fri, 28 Mar 2014 17:52:09 +0000 (17:52 +0000)]
process stack command even if gatekeeper client isn't register
don't destroy gatekeeper client if it is not started
don't destroy gatekeeper client in some sort of gatekeeper errors
signal rtp create condition when call cleared before rtp structure created
Matthew Jordan [Fri, 28 Mar 2014 17:35:48 +0000 (17:35 +0000)]
Update API versions and UPGRADE/CHANGES for 12.2.0
This patch does the following:
* It updates the AMI version to 2.2.0 to indicate backwards compatible
changes have been made since the last release
* It updates the ARI version to 1.2.0 to indicate backwards compatible
changes have been made since the last release
* It updates the UPGRADE/CHANGES files with changes that were not
mentioned
Matthew Jordan [Fri, 28 Mar 2014 16:48:32 +0000 (16:48 +0000)]
res_config_odbc/res_odbc: Fix handling of non-text columns updates with empty values.
This patch fixes setting nullable integer columns to NULL instead of an empty
string, which fails for PostgreSQL, for example. The current code is supposed
to do so, but the check is broken. The patch also allows the first column in
the list to be a nullable integer.
This patch also adds a compatibility setting in res_odbc.conf,
allow_empty_string_in_nontext. It is enabled by default. It should be disabled
for database backends (such as PostgreSQL) that require NULL instead of an
empty string for Integer columns.
http: response body often missing after specific request
This patch works around a problem with the HTTP body
being dropped from the response to a specific client
and under specific circumstances:
a) Client request comes from node.js user agent
"Shred" via use of swagger-client library.
b) Asterisk and Client are *not* on the same
host or TCP/IP stack
In testing this problem, it has been determined that
the write of the HTTP body is lost, even if the data
is written using low level write function. The only
solution found is to instruct the TCP stack with the
shutdown function to flush the last write and finish
the transmission. See review for more details.
Matthew Jordan [Fri, 28 Mar 2014 14:18:32 +0000 (14:18 +0000)]
contrib/realtime: Remove empty SQL script files
Since the relatime scripts are now managed by Alembic, the previous realtime
scripts were previously removed. However, the removal process messed up, as
the files were still in the repository. The contents were just empty.
projects / thirdparty / asterisk.git / log
