FS-7083 #resolve #comment this should do it.  The problem is linked to side-effects from the read thread being delayed by writing to the file handle.  It was so much worse on mp3 because the shout encoder blocks while its churning the data and delays it more.  This patch adds a dedicated thread for writing to the file and the channel_variable RECORD_USE_THREAD=false will disable it and sync may still be maintained at the cost of dropping more data from the audio signal.

* commit 'c96d64e7':
  invalid type cast correction
  list_users tipo correction

Merge pull request #143 in FS/freeswitch from bugfix/FS-7095-freeswitch-sends-dtls-hello-and-stun to master

* commit 'ba016c28':
  FS-7095 #comment please test

FS-7083 #comment patch to change mod_shout to use lame_encode_buffer_interleaved on stereo channels so we don't have to mess with the input data

FS-7091 #resolve #comment [mod_rayo] Removed unnecessary mutex lock inside input component's cleanup function since the input component won't be cleaned up unless all references have been released.  This fixes a deadlock when output component is created at the same time an input component completes.

The following values can be read from the event:

sync_lost_percent - Error percentage within the analysis window
sync_lost_count - How many times sync has been lost
cng_count - Counter of confort noise packets
err_samples - Number of samples that did not match the sequence

* Use memmem() and memcmp() for pattern checks
* Use a static pre-built buffer for the write frame

windows force datatype conversion on uint64_t to switch_size_t for recent change. maybe we should change the structure to be uint64_t as well?

Merge pull request #127 in FS/freeswitch from ~ITALOROSSI/freeswitch:mod_callcenter-queue-list-count-by-state to master

* commit 'f63f8686':
  FS-7049 - Documentation for state optional paramenter in callcenter_config queue list and count
  fix FS-7049 - Count and list agents based on their state

re-mark cur_payload as negotiated when detected as such by parser or the rtp could stop working on session re-invite

FS-7015 #resolve #comment The code was not properly catching the 0.0.0.0 after changing it to work with ICE SDPs because it was looking in the wrong place for the 0.0.0.0

* commit '1bcdb49b':
  Fix JIRA (FS-7018): mod_callcenter's longest-idle-agent strategy is unfair

only negotiate codecs that were previously negotiated on re-invites unless its intentional to change codecs

FS-6891 FS-7002 FS-7059 FS-7072 FS-7073 FS-7076 #close #comment All of these bugs are invalidated due to a botched revert -- commit 72c3df5e was supposed to revert commit 1b612fec but it only reverted part of it.  commit 158c1f23 from the other day was erroneously created to work around the result of the botched revert so it is reverted.  The result of this commit should be to bring it back to effectively properly revert 1b612fec and now you can test and open new bugs if necessary.  This includes the segfault in eavesdrop, and any media bug releated reports.  Consider working together and only reporting one new bug if you still encounter any more problems and be sure to include details and make sure the jitter buffer is being used.

FS-7047 #resolve #comment [mod_rayo] arbitrary MRCP headers can now be sent to unimrcp output components

Merge pull request #139 in FS/freeswitch from bugfix/FS-6688-notify-message-record-route-tag to master

* commit '5376e826':
  FS-6688: don't update the contact when the original had fs_path, the new one doesnt, and the contact didn't change.  This will fix the normal case of record route from a proxy without breaking normal changing of a contact

FS-6688: don't update the contact when the original had fs_path, the new one doesnt, and the contact didn't change.  This will fix the normal case of record route from a proxy without breaking normal changing of a contact

Merge pull request #137 in FS/freeswitch from bugfix/FS-7061-add-q-header-param-on-300-responses to master

* commit '21458f85':
  FS-7062: [mod_sofia] on redirect, when uri are passed in without <> with multiple uris, automatically add the q= header param in decending order.  This should make 300 Multiple Choices work well with devices that require the q param.  If you would like to specify explicit q-values, please use the syntax of redirect where you specify the entire header using the <>