It's broken too many of my recent builds.  We can add it back in when
things stabilize or perhaps after we get it to swig with a recent
version of swig that we can make part of the clean bootstrap/build
process.

POSIX sed doesn't support grouping, so split the regex in two.

Reference-URL: http://jira.freeswitch.org/browse/FS-4276Signed-off-by: Stefan Knoblich <stkn@openisdn.net>

Signed-off-by: Stefan Knoblich <stkn@openisdn.net>

Speculative fix for a mod_python build issue on OSX, which is
caused by '-arch i386 -arch x86_64' in the final PYTHON_CFLAGS.

Filter both flags and hope OSX's universal binary toolchain is smart enough
to choose a sane default.

Reported-by: neilp9 in #freeswitch @ irc.freenode.net
Signed-off-by: Stefan Knoblich <stkn@openisdn.net>

Our default build probably shouldn't include non-free software.  With
mod_ilbc, the licensing situation is merely ambiguous.  With
mod_siren, the user can't use this code without getting explicit
permission from Polycom (though it is apparently easily given).

The test was backward.  If the revision is in switch_version.h.in then
we have the version already.

These codecs are non-free which creates issues for distributions, so
let's not require these by default to run our example configs.  We can
add back in iLBC once we resolve the licensing situation with our
in-tree implementation.

We believe it to be safe now.  In test run on multi-core machines it
is producing bit-for-bit identical result compared to the non-parallel
invocation.

build_all call each of these subroutines under a sub-shell, so the set
-e gets wiped out.

Because we're capturing the string return value of these functions, we
can't see any output that goes to stdout.