Merge pull request #1335 in FS/freeswitch from ~ANDYWOLK/freeswitch:FS-10489-move-libg7221-from-tree-to-download to master

* commit '06ef6310':
  FS-10489: [build-system] Move libg7221 from tree to download on Windows.

spandsp DSP tweaks to avoid problems with the LPC10 code compiled with the latest GCC 6.2 series compilers. Various other additional tweaks to spandsp.

FS-9979: [verto.js] When verto client rejects a call returns hangup cause NORMAL_CLEARING #comment passing neither cause or causeCode will result in normal clearing, passing either one of cause or causeCode will result in that value being passed and parsed on the server side, if the cause is valid it will be accepted otherwise it will try to validate causeCode instead, if neither are true it defaults to normal clearing #resolve

FS-10495: [Modules] firing custom event with no sub class will crash freeswitch when mod_fail2ban is loaded #resolve

This commit causes a segfault when parsing destinations that are |
delimited. This patchset needs to be tested further before inclusion
into the tree.

This reverts commit b11955db.

* commit '999a6dd0':
  FS-10485: Re-activate ICE on all attribute changes

Merge pull request #1329 in FS/freeswitch from ~ANDYWOLK/freeswitch:bugfix/FS-10488-fix-memory-leak to master

* commit '28913d27':
  FS-10488: [mod_v8] Fix memory leak.

Merge pull request #1324 in FS/freeswitch from ~ANDYWOLK/freeswitch:bugfix/FS-10465-fix-javascript-failures-when-a-session to master

* commit '105afc3f':
  FS-10465: [mod_v8] Fix javascript failures when a session callback is called from another thread/leg.

I ran into some ice failures when testing reinvites using
bypass_media_after_bridge. Every ice session has ice-ufrag and ice-pwd
attributes. When leg A sends those attributes to leg B, leg B is
supposed to send stun packets to leg A using the credentials provided in
those attributes. Freeswitch was not updating the session credentials
when the phone sent a new invite containing new ufrag and pwd
attributes. This caused ICE to fail on the client side (PJSIP in this
case).

The new_ice flag is set to true when freeswitch detects a change in
those attributes. For whatever reason, the reinvite flows that I was
testing did not set the CF_REINVITE flag on the channel, but new_ice was
true. It seems to me that freeswitch should always reactivate ice when
the ufrag and pwd credentials change.

Update Debian Package Builds in prep for early early debian 9 packages and for changes in the master branch.

Clears SSL error state.

Updates tport_tls.c to clear the SSL error state after an error occurs.

OpenSSL puts errors into a queue that is kept in thread local storage.
In some cases, such as when SSL_ERROR_SSL is returned by SSL_get_errror(),
OpenSSL will queue multiple errors for a single event. When this occurs,
OpenSSL will report an error the next time I/O is performed if the queue
is not cleared first, which can result in TLS connections being torn down
prematurely.

Merge pull request #1306 in FS/freeswitch from ~HRISTO/freeswitch:FS-10407-set-some-redirect-channel-variables to master

* commit '1d15e411':
  FS-10407: [mod_sofia] Set redirect variables when outbound_redirect_fatal is true

Merge pull request #1313 in FS/freeswitch from ~ANTONIO/freeswitch:bugfix/FS-10368-mod_callcenter-lock-database-errors to master

* commit '04854fdd':
  Revert "FS-10299 [mod_callcenter] Removing global lock on all cc_execute_sql functions when executing database queries"

Merge pull request #1307 in FS/freeswitch from ~HRISTO/freeswitch:feature/FS-10416-add-new-endpoints-to-be-called-while to master

* commit 'b11955db':
  FS-10416: [mod_commands] add new outbound channels to an in-progress originate

* commit 'd5cceb30':
  FS-10461: created scripts to build docker base image

* commit '0fc8c71a':
  FS-10451: Updated sound files descriptions

* commit 'aaa5f0f6':
  FS-10452: Fixed build of docker container

FS-10473: [freeswitch-core] FreeSWITCH crash - Null event pointer dereference during conference_cdr_del #resolve

FS-10286: [mod_conference] Sync member joins up with keyframes in shared encoder mode -- high cpu usage on h264

Merge pull request #1273 in FS/freeswitch from ~EDUARDO/freeswitch:FS-10304-prevent-infinite-logging-when-a to master

* commit 'be741eab':
  FS-10304: [mod_callcenter] Prevent infinite logging when a stale queue member in found in the database

FS-10430: [mod_amqp] fix support for libamqp version .0.5.2.  Change to api actually happend in v0.6.0

FS-10465: [mod_v8] Fix javascript failures when a session callback is called from another thread/leg.