Steve Murphy [Tue, 2 Sep 2008 22:40:13 +0000 (22:40 +0000)]
After reconsidering, with respect to 13409, ast_cdr_detach should be OK, better in fact, than ast_cdr_free, which generates lots of useless warnings that will undoubtably generate complaints.
I basically spent the day, verifying that this patch
solves the problem, and doesn't hurt in non-problem
cases. Why valgrind did not plainly reveal this leak
absolutely mystifies and stuns me.
Many, many thanks to tomaso for finding and providing the fix.
Mark Michelson [Fri, 29 Aug 2008 17:34:17 +0000 (17:34 +0000)]
After working on the ao2_containers branch, I noticed
something a bit strange. In all cases where we provide
a callback function to ao2_container_alloc, the callback
function would only return 0 or CMP_MATCH. After inspecting
the ao2_callback() code carefully, I found that if you're
only looking for one specific item, then you should return
CMP_MATCH | CMP_STOP. Otherwise, astobj2 will continue
traversing the current bucket until the end searching for
more matches.
In cases like chan_iax2 where in 1.4, all the peers are
shoved into a single bucket, this makes for potentially
terrible performance since the entire bucket will be
traversed even if the peer is one of the first ones come
across in the bucket.
All the changes I have made were for cases where the
callback function defined was passed to ao2_container_alloc
so that calls to ao2_find could find a unique instance
of whatever object was being stored in the container.
Mark Michelson [Fri, 29 Aug 2008 16:01:07 +0000 (16:01 +0000)]
Add context checking when retrieving a vm_state.
This was causing a problem for people who had identically
named mailboxes in separate voicemail contexts.
This commit affects IMAP storage only.
Mark Michelson [Wed, 27 Aug 2008 19:49:20 +0000 (19:49 +0000)]
Fix tag checking in get_sip_pvt_byid_locked when
in pedantic mode. The problem was that the wrong
tags would be compared depending on the direction
of the call.
Russell Bryant [Tue, 26 Aug 2008 16:07:58 +0000 (16:07 +0000)]
Fix some bogus scheduler usage in chan_sip. This code used the return value
of a completely unrelated function to determine whether the scheduler should
be run or not. This would have caused the scheduler to not run in cases where
it should have. Also, leave a note about another scheduler issue that needs
to be addressed at some point.
Russell Bryant [Tue, 26 Aug 2008 15:27:23 +0000 (15:27 +0000)]
Fix a race condition with the IAX scheduler thread. A lock and condition are
used here to allow newly scheduled tasks to wake up the scheduler just in case
the new task needs to run sooner than the current wakeup time when the thread
is sleeping. However, there was a race condition such that a newly scheduled
task would not properly wake up the scheduler or affect the wake up period.
The order of execution would have been:
1) Scheduler thread determines wake up time of N ms.
2) Another thread schedules a task and signals the condition, with an
execution time of < N ms.
3) Scheduler thread locks and goes to sleep for N ms.
By moving the sleep time determination to inside the critical section, this
possibility is avoided.
Terry Wilson [Tue, 26 Aug 2008 15:22:58 +0000 (15:22 +0000)]
sounds/Makefile installs sounds using the "new" language directory structure, but languageprefix needs to be set = yes for sounds in subdirectories (digits/1, etc.) to play as the correct language.
Fix the generation of asterisk.conf to include languageprefix=yes
Mark Michelson [Mon, 25 Aug 2008 15:52:59 +0000 (15:52 +0000)]
Fix the logic in config_text_file_save so that if an
UpdateConfig manager action is issued and the
file specified in DstFileName does not yet exist,
an error is not returned.
(closes issue #13341)
Reported by: vadim
Patches:
13341.patch uploaded by putnopvut (license 60)
(with small modification from seanbright)
Steve Murphy [Fri, 22 Aug 2008 22:24:02 +0000 (22:24 +0000)]
I found some problems with the code I committed earlier, when
I merged them into trunk, so I'm coming back to clean up.
And, in the process, I found an error in the code I added
to trunk and 1.6.x, that I'll fix using this patch also.
Mark Michelson [Fri, 22 Aug 2008 19:23:02 +0000 (19:23 +0000)]
Remove show_frame_stats_deprecated since it is not
used anywhere and causes build errors if building under
dev-mode with TRACE_FRAMES selected in menuselect.
THis is a bold move for a static release fix, but I wouldn't have
made it if I didn't feel confident (at least a *bit* confident)
that it wouldn't mess everyone up.
The reasoning goes something like this:
1. We simply cannot do anything with CDR's at the current point
(in pbx.c, after the __ast_pbx_run loop). It's way too late to
have any affect on the CDRs. The CDR is already posted and gone,
and the remnants have been cleared.
2. I was very much afraid that moving the running of the 'h'
extension down into the bridge code (where it would be now
practical to do it), would result in a lot more calls to the
'h' exten, so I implemented it as another exten under another
name, but found, to my pleasant surprise, that there was a
1:1 correspondence to the running of the 'h' exten in the
pbx_run loop, and the new spot at the end of the bridge.
So, I ifdef'd out the current 'h' loop, and moved it into
the bridge code. The only difference I can see is the stuff
about the AST_PBX_KEEPALIVE, and hopefully, if this
is still an important decision point, I can replicate it
if there are complaints. To be perfectly honest,
the KEEPALIVE situation is not totally clear to me,
and how it relates to a post-bridge situation is less
clear. I suspect the users will point out everything
in total clarity if this steps on anyone's toes!
3. I temporarily swap the bridge_cdr into the channel
before running the 'h' exten, which makes it possible
for users to edit the cdr before it goes out the door.
And, of course, with the endbeforehexten config var set,
the users can also get at the billsec/duration vals.
After the h exten finishes, the cdr is swapped back
and processing continues as normal.
Please, all who deal with CDR's, please test this version
of Asterisk, and file bug reports as appropriate!
Russell Bryant [Wed, 20 Aug 2008 22:14:35 +0000 (22:14 +0000)]
Fix a crash in the ChanSpy application. The issue here is that if you call
ChanSpy and specify a spy group, and sit in the application long enough looping
through the channel list, you will eventually run out of stack space and the
application with exit with a seg fault. The backtrace was always inside of
a harmless snprintf() call, so it was tricky to track down. However, it turned
out that the call to snprintf() was just the biggest stack consumer in this
code path, so it would always be the first one to hit the boundary.
Kevin P. Fleming [Wed, 20 Aug 2008 19:35:59 +0000 (19:35 +0000)]
Backport support for Zaptel/DAHDI channel-level alarms from trunk/1.6, because not doing so just makes it difficult for people with channels that are in alarm when Asterisk starts up to get them going once the alarm is cleared
The specialized reset routine is tromping on the
flags field of the CDR. I made a change to not
reset the DISABLED bit. This should get rid of this
problem.
Mark Michelson [Tue, 19 Aug 2008 18:50:53 +0000 (18:50 +0000)]
Add a lock and unlock prior to the destruction of the chanspy_ds
lock to ensure that no other threads still have it locked. While
this should not happen under normal circumstances, it appears that
if the spyer and spyee hang up at nearly the same time, the following
may occur.
1. ast_channel_free is called on the spyee's channel.
2. The chanspy datastore is removed from the spyee's channel in
ast_channel_free.
3. In the spyer's thread, the spyer attempts to remove and destroy the datastore
from the spyee channel, but the datastore has already been removed in step 2,
so the spyer continues in the code.
4. The spyee's thread continues and calls the datastore's destroy callback,
chanspy_ds_destroy. This involves locking the chanspy_ds.
5. Now the spyer attempts to destroy the chanspy_ds lock. The problem is that in step 4,
the spyee has locked this lock, meaning that the spyer is attempting to destroy a lock
which is currently locked by another thread.
The backtrace provided in issue #12969 supports the idea that this is possible
(and has even occurred). This commit does not close the issue, but should help
in preventing one type of crash associated with the use of app_chanspy.
Jeff Peeler [Fri, 15 Aug 2008 19:21:51 +0000 (19:21 +0000)]
Fixes the dahdi restart functionality. Dahdi restart allows one to restart all DAHDI channels, even if they are currently in use. This is different from unloading and then loading the module since unloading requires the use count to be zero. Reloading the module is different in that the signalling is not changed from what it was originally configured. Also, this fixes not closing all the file descriptors for D-channels upon module unload (which would prevent loading the module afterwards).
Russell Bryant [Fri, 15 Aug 2008 15:07:16 +0000 (15:07 +0000)]
Ensure that when a hangup occurs in autoservice, that a hangup frame gets
properly deferred to be read from the channel owner when it gets taken out
of autoservice.
Tilghman Lesher [Thu, 14 Aug 2008 16:47:30 +0000 (16:47 +0000)]
When creating the secondary subchannel name, it is necessary to compare to
the existing channel name without the "Zap/" or "DAHDI/" prefix, since our
test string is also without that prefix.
(closes issue #13027)
Reported by: dferrer
Patches:
chan_zap-1.4.21.1_fix2.patch uploaded by dferrer (license 525)
(Slightly modified by me, to compensate for both names)
Russell Bryant [Thu, 14 Aug 2008 14:05:23 +0000 (14:05 +0000)]
Comments in this config file were aligned only if your tab size was set to 8.
So, convert tabs to spaces so that things should be aligned regardless of what
tab size you use in your editor.
Sean Bright [Wed, 13 Aug 2008 00:10:35 +0000 (00:10 +0000)]
Bring cdr_tds in line with the other CDR backends and have it try to store
CDR(userfield) if it is set. The new behavior is to check for the userfield
column on module load, and if it exists, we will store CDR(userfield) when
CDRs are written. A similar patch already went into trunk and 1.6.0.
Steve Murphy [Fri, 8 Aug 2008 00:15:34 +0000 (00:15 +0000)]
(closes issue #13236)
Reported by: korihor
Wow, this one was a challenge!
I regrouped and ran a new strategy for
setting the ~~MACRO~~ value; I set it once
per extension, up near the top. It is only
set if there is a switch in the extension.
So, I had to put in a chunk of code to detect
a switch in the pval tree.
I moved the code to insert the set of ~~exten~~
up to the beginning of the gen_prios routine,
instead of down in the switch code.
I learned that I have to push the detection
of the switches down into the code, so everywhere
I create a new exten in gen_prios, I make sure
to pass onto it the values of the mother_exten
first, and the exten next.
I had to add a couple fields to the exten
struct to accomplish this, in the ael_structs.h
file. The checked field makes it so we don't
repeat the switch search if it's been done.
Shaun Ruffell [Thu, 7 Aug 2008 18:14:17 +0000 (18:14 +0000)]
Updated codec_dahdi to use the new transcoder interface in the first DAHDI
release. Codec dahdi no longer functions with the transcoder interface in
zaptel at this time (which the last zaptel release was 1.4.11).
NOTE: Still needs an update to the configure script to make sure that
codec_dahdi is only built if the new transcoder interface is present in the
drivers.
Russell Bryant [Wed, 6 Aug 2008 18:58:37 +0000 (18:58 +0000)]
Remove the use of --no-print-directory when compiling subdirectories.
This allows vim :make functionality to work properly when errors have
occurred in the build. Without printing the directories, vim did not
know how to find the file that the error occurred in.
If the extra bit of build noise annoys anyone, just let me know, and
I'll make this optional.
Mark Michelson [Wed, 6 Aug 2008 15:58:40 +0000 (15:58 +0000)]
Since adding the AST_CONTROL_SRCUPDATE frame type,
there are places where ast_rtp_new_source may be called
where the tech_pvt of a channel may not yet have an
rtp structure allocated. This caused a crash in chan_skinny,
which was fixed earlier, but now the same crash has been
reported against chan_h323 as well. It seems that the best
solution is to modify ast_rtp_new_source to not attempt to
set the marker bit if the rtp structure passed in is NULL.
This change to ast_rtp_new_source also allows the removal
of what is now a redundant pointer check from chan_skinny.
Mark Michelson [Wed, 6 Aug 2008 00:25:10 +0000 (00:25 +0000)]
Merging the issue11259 branch.
The purpose of this branch was to take into account
"burps" which could cause jitterbuffers to misbehave.
One such example is if the L option to Dial() were used
to inject audio into a bridged conversation at regular
intervals. Since the audio here was not passed through
the jitterbuffer, it would cause a gap in the jitterbuffer's
timestamps which would cause a frames to be dropped for a
brief period.
Now ast_generic_bridge will empty and reset the jitterbuffer
each time it is called. This causes injected audio to be handled
properly.
ast_generic_bridge also will empty and reset the jitterbuffer
if it receives an AST_CONTROL_SRCUPDATE frame since the change
in audio source could negatively affect the jitterbuffer.
All of this was made possible by adding a new public API call
to the abstract_jb called ast_jb_empty_and_reset.
I discovered that also, in the previous bug fixes and changes,
the cdr.conf 'unanswered' option is not being obeyed, so
I fixed this.
And, yes, there are two 'answer' times involved in this
scenario, and I would agree with you, that the first
answer time is the time that should appear in the CDR.
(the second 'answer' time is the time that the bridge
was begun).
I made the necessary adjustments, recording the first
answer time into the peer cdr, and then using that to
override the bridge cdr's value.
To get the 'unanswered' CDRs to appear, I purposely
output them, using the dial cmd to mark them as
DIALED (with a new flag), and outputting them if
they bear that flag, and you are in the right mode.
I also corrected one small mention of the Zap device
to equally consider the dahdi device.
I heavily tested 10-sec-wait macros in dial, and
without the macro call; I tested hangups while the
macro was running vs. letting the macro complete
and the bridge form. Looks OK. Removed all the
instrumentation and debug.
Tilghman Lesher [Tue, 5 Aug 2008 21:34:46 +0000 (21:34 +0000)]
In a conversion to use ast_strlen_zero, the meaning of the flag IAX_HASCALLERID
was perverted. This change reverts IAX2 to the original meaning, which was,
that the callerid set on the client should be overridden on the server, even if
that means the resulting callerid is blank. In other words, if you set
"callerid=" in the IAX config, then the callerid should be overridden to blank,
even if set on the client. Note that there's a distinction, even on realtime,
between the field not existing (NULL in databases) and the field existing, but
set to blank (override callerid to blank).
Specify codecs in callfiles and manager, to allow video calls to be set up
from callfiles and AMI.
(closes issue #9531)
Reported by: Geisj
Patches:
20080715__bug9531__1.4.diff.txt uploaded by Corydon76 (license 14)
20080715__bug9531__1.6.0.diff.txt uploaded by Corydon76 (license 14)
Tested by: Corydon76
OK, a few days of debugging, a bunch of instrumentation
in chan_sip, main/channel.c, main/pbx.c, etc. and 5 solid
notebook pages of notes later, I have made the small
tweek necc. to get the start time right on the second
CDR when:
A Calls B
B answ.
A hits Xfer button on sip phone,
A dials C and hits the OK button,
A hangs up
C answers ringing phone
B and C converse
B and/or C hangs up
But does not harm the scenario where:
A Calls B
B answ.
B hits xfer button on sip phone,
B dials C and hits the OK button,
B hangs up
C answers ringing phone
A and C converse
A and/or C hangs up
The difference in start times on the second CDR is because
of a Masquerade on the B channel when the xfer number is
sent. It ends up replacing the CDR on the B channel with
a duplicate, which ends up getting tossed out. We keep
a pointer to the first CDR, and update *that* after the
bridge closes. But, only if the CDR has changed.
I hope this change is specific enough not to muck
up any current CDR-based apps. In my defence, I
assert that the previous information was wrong,
and this change fixes it, and possibly other
similar scenarios.
I wonder if I should be doing the same thing
for the channel, as I did for the peer, but
I can't think of a scenario this might affect.
I leave it, then, as an exersize for the users,
to find the scenario where the chan's CDR
changes and loses the proper start time.
Mark Michelson [Thu, 31 Jul 2008 15:56:18 +0000 (15:56 +0000)]
Add more timeout checks into app_queue, specifically
targeting areas where an unknown and potentially
long time has just elapsed. Also added a check
to try_calling() to return early if the timeout
has elapsed instead of potentially setting a negative
timeout for the call (thus making it have *no* timeout
at all).
projects / thirdparty / asterisk.git / log
