Apparently OS X ships a sed that requires an argument to -i.

In UNIX, text files by definition end with a newline.

write(3) can write fewer bytes than was requested for any number of
reasons.  The correct behavior is to retry unless there is an error.

If there is an error, try to unlink the file; no sense in leaving
corrupted data laying around.

The default value of libdir is (unexpanded) '${exec_prefix}/lib'.  In
the non-FHS path this is fine because it only ends up in a variable
where it will be expanded later.  By using this to define modulesdir
we let it slip into a define where it made no sense.

Order the default configure flags in the order they show up in
configure --help, and order our own alphabetically.

When --enable-fhs is passed to configure, we set all paths by default
in a way compliant with FHS, the Filesystem Hierarchy Standard.

  http://www.pathname.com/fhs/

Each path may still be overridden by passing the specific flag for it.

This shows the cipher name, TLS version, the number of cipher bits and
algorithm bits, and a description of the cipher in Sofia's debug
logging output on level 9.

Unlike fread(3), read(3) will return -1 on error.  We were assigning
the result of read to a potentially unsigned variable, and passing the
result down to switch_xml_parse_str() where it would end up
determining how many bytes to malloc(3).

This looks like a copy/paste error.

FS-6037 --resolve

Thanks-to: David Sanders <dsanders@pinger.com>

This man page is written in ronn format (package ruby-ronn on Debian)
which is a subset of markdown that can be converted to nroff man
pages, HTML, and other formats.

We want people to have this on so that when we're looking at reported
log files we can tell what is going on.

This avoids searching the string repeatedly with strchr.

We were incorrectly parsing usernames and domains starting with "sip"
if there was no sip: or sips: scheme in the string.

We were also incorrectly parsing usernames containing a colon even if
a scheme was given.

This also refactors the function for hopefully greater clarity.

In commit 7efeabbd Anthony fixed the
handling of sip:example.com and sips:example.com URLs, however he
introduced a regression causing URLs starting with 's' to be parsed
incorrectly.

In commit 7d2456ea Brian fixed the
regression, but introduced a regression causing sips:example.com URLs
to be handled incorrectly.

FS-6226Prevent DTMF from traversing bridged channels, but still allow me to send DTMF via API or dp app (uuid_send_dtmf or send_dtmf)

Originally we did the same thing with SRTP that we do without SRTP,
which is to simply not send packets when e.g. sleep is called.

At commits d6332397 and
5259814a we enabled sending silence
packets with comfort noise when SRTP is active.  We appear to have
done this for interop purposes; many devices can't handle gaps in the
stream of SRTP packets.

But our current comfort noise implementation doesn't take the codec
rate into account (FS-6291), so on 16kHz codecs the constant we chose
created an annoying level of static between sound file playback.

With this commit we preserve the sending of SRTP packets during idle
periods, but make those packets completely silent.

Thanks-to: Anthony Minessale <anthm@freeswitch.org>

FS-5053 --resolve

Conflicts:
	src/mod/applications/mod_conference/mod_conference.c

Conflicts:
	src/mod/applications/mod_conference/mod_conference.c

FS-5886 --resolve add NDLB_broken_opus_sdp which must be set as a global variable in var.xml or switch.conf.xml to enable broken opus mode for back compat with counterpath mobile

fix race condition where a transferring leg could be hungup on by the bridge partner from the previous bridge because of hangup held leg detection.  The leg which was hungup when held doesn't realize the other leg has already moved on from the bridge because it was transferred and is already on its way to connect to the new destination

Our sqlite fork was reading past the end of the zP3 KeyInfo structure
here.  This was causing gcc and clang's address sanitization to alert.

FS-6279 --resolve

Thanks-to: Christopher Rienzo <chris@rienzo.net>

FS-5934 force_transfer_context not being honored if you happen to do an attended this patch makes perfect sense.

Conflicts:
	src/include/switch_types.h

FS-6250 fix regression from 7efeabbd, don't discrimiate against people that have s as the first character in their user names

FS-5888 --resolve Rather than changing the default behavior I've updated the log line to be more accurate. Also fixed the problem where the delay was being overwritten.

Pre-jessie versions of Debian's mk-build-deps tool rely on libraries
that do not correctly parse debian/control, so we have to clip it down
to a smaller subset of the format to build on e.g. wheezy.