Do not perform protocol negotiations in an offer
closes#1086
Change-Id: I5fb7638aa71e80c08abcda2d8c341ae72725bace
(cherry picked from commit 92ea2685ae)
An endpoint switching protocols is normally passed through to the peer,
but in an answer that is usually not desirable. Change the default
behaviour to stick to the original protocol that was offered even if the
answerer changes protocols.
Change-Id: Ib288549f4b1c9ab57a6333c6b7dd511537af96f7
(cherry picked from commit 12ac8846de)
This is needed to reset the kernel stream indexes.
fixes#1048
Change-Id: If46c4f85a752c2ed64997e6c4a44625845f4cc87
(cherry picked from commit 214f30e8cf)
The previous approach causes problems during re-invites with updated
port (or address) information. The UA might delay actually using the new
endpoint until after the answer is seen, in which case rtpengine would
erroneously immediately latch on to the old endpoint. We need to
reinstate the endpoint learning delay in all cases, and make sure
previously learned endpoints are forgotten whenever a new signal is
seen.
Change-Id: If42ed13c43710616c02cdf018b36a4543d7f6bfe
(cherry picked from commit 112e57f9fe)
Initially created branches don't have a tag associated with them, so
they don't appear in `other_tags`. We need to keep track of which
branches were created from a single from-tag through their via-branch
values so we know when all branches have been deleted.
closes#1037
Change-Id: I4c6ce602b99a9104de98098cd06ca399eb28e1b2
* use auto cleanup buffer variable
* use g_vasprintf instead of libc vasprintf
* rename length variable for clarity
Change-Id: Ibebb8cccd51aa2f5cf3704b73bf82472b73d6f19
On some systems (e.g. Github Travis) make seems to have a problem with
its job control, resulting in a race condition where multiple make
instances try to build the same file at the same time. For most files
this is not a problem, but those auto-generated using an append
operation (>>) the output file can become garbled. Solve this by using
an atomic file creation method.
Change-Id: I1acb2387e2ca6ca30ddc3742143fd910032bfbb0
Commit a6d2f9296 introduced usage of libevent_global_shutdown(),
which is available in libevent-dev >=2.1.1-alpha only.
For example travis CI is running Ubuntu 16.04 / xenial and ships
libevent-dev 2.0.21-stable-2, so we can't just raise the
build-dependency of libevent-dev to >= 2.1.1+ if we still want
to support older systems.
Change-Id: I9dcf9ca0a19f59051d42f0d2383d1938b60a97df
iptables-dev is only available until Debian/buster and no longer exists in
Debian/bullseye (current testing) nor unstable/sid:
| builddeps:. : Depends: iptables-dev (>= 1.4) but it is not installable
Even in Debian/buster iptables-dev is already a transitional/dummy package.
Support iptables-dev as alternative Build-Dependency, just in case
someone is building the package against a system where libxtables-dev
doesn't exist yet.
Change-Id: I28c4c81ac474c646d80a0146baa2446dde7073c3
The file utils/build_deps.sh was removed in commit e580c7c7f8,
adjust shellcheck execution accordingly.
Change-Id: Ibe454e98d02e026ef5cade9401e73a04975e6cc3
shellcheck v0.7.1 complains about a bunch of issues:
SC2154: status is referenced but not assigned.
SC2181: Check exit code directly with e.g. 'if mycmd;', not indirectly with $?.
SC2207: Prefer mapfile or read -a to split command output (or quote to avoid splitting).
SC2235: Use { ..; } instead of (..) to avoid subshell overhead.
SC2236: Use -n instead of ! -z.
The "$status" variable disappeared in d4763aba14.
The init scripts of ngcp-rtpengine-daemon and
ngcp-rtpengine-recording-daemon had a logic bug, where a failing stop
action didn't properly return, but continued execution of the following
firewalling code (ngcp-rtpengine-iptables-setup). Thanks to Guillem
Gover for spotting this one. While at it, no longer execute under
'set -e'.
Change-Id: Ia50e76f615564a288627e6e42ec8f7eb082de74c
Richard Fuchs
Sipwise Jenkins Builder
Stefan Mititelu
Michael Prokop