But use old jose 1.11.1 on Erlang/OTP older than 21
Erlang older than 21 does not support
try catch _Class:_Reason:_Stacktrace ->
that is used in jose_server.erl
This gives a clear definition of what erlang versions require each
workaround, instead of depending on the actual workaround details.
The lib module was removed in OTP 21.0,
and the formatting options were moved to erl_error.
The uri_string module was introduced in OTP 21.0.
db_type as a compilation define was first introduced in 2006 in d0e2ac9,
but was removed in 2015 in 25791cf.
Nowadays: ./configure --enable-mssql will set {mssql, true} in vars.config;
That is read by ejabberd.app.src.script to set enabled_backends;
which is consulted by test/suite.erl to test only available backends.
Prepare with:
./autogen.sh && ./configure --with-rebar=./rebar3 && make
Or use this if you installed Elixir:
./autogen.sh && ./configure --with-rebar=mix && make
Start without installing (it recompiles when necessary):
make relive
It stores config, database and logs in _build/relive/
There's available the well-known script:
_build/relive/ejabberdctl
Please note this fails immediately:
r3:do(compile).
This crashes a few seconds later:
rebar3:run(["compile"]).
Workaround that works correctly:
ejabberd_admin:update().
Rebar 2 expands {{erts-vsn}} to "erts-$vsn", Rebar 3 expands it to just
"$vsn". Make sure `make rel` doesn't end up with a "$vsn" directory
next to "erts-$vsn" (which happened when using Rebar 3), and make sure
that ejabberdctl expects both "erl" and "epmd" to be installed below
"erts-$vsn" (which it didn't when using Rebar 3).
That script serves a similar purpose to ejabberdctl to start ejabberd,
but we can't guarantee it is completely equivalent to ejabberdctl.
The prod profile must provide only the well-known script.
The test profile provides the Relx script so we can experiment with it.
This reverts commit 0881c5941a.
The extended start had been enabled intentionally, as some commands
(such as 'ping', 'status', or 'pid') actually do work.
Flock'ing /var/lock/ejabberdctl by ejabberdctl was removed with
f7d4aae64d ("Use UUID for ctl node name (#1021)"), however the
according recipies in the Makefile where never removed. This commit
does that.
Let clients request the maximum limit for the node configuration option
'max_items' by specifying the special value 'max' instead of an integer.
This was added to XEP-0060, revision 1.17.0 (and clarified in revision
1.20.0).
Thanks to Ammonit Measurement GmbH for sponsoring this work.
Notice that Elixir 1.12.0 requires Erlang/OTP 21 or higher
BTW, this workaround works only with rebar2... unfortunately
rebar3 doesn't compile the elixir files from lib/
That's the only way to ensure that tar.gz will contain include/
of ejabberd and its dependencies, which is required to compile
ejabberd-contrib modules with "ejabberdctl module_install ..."
It supports all that was already in .travis.yml and tests.yml, plus:
- exclude paths that don't affect source code
- two very distinct scenarios, ancient a recent (OTP, Rebar, Ubuntu)
- cache binaries and dialyzer for small speedup and network usage
- separated results, to view comfortably where and what failed
- use coveralls example rebar3 code, instead of custom rebar2
- upload binaries.zip that can be used with Binary Installers
Reported in #3527. Add hostname matching function, and specify SNI
Also, OTP 23 dropped backwards compatibility for 0, 1, 2 values for verify, so
replace with combination of verify_none/verify_peer and fail_if_no_peer_cert
as appropriate
rebar3 needs sql directory as a src dir so that tests can reference sql
files, so sql dir was added to extra_src_dirs in d9c1befb. It turns out
extra_src_dirs does unexpected unwanted extra things like copy all beam
files there too, so move sql dir to regular src_dirs
erlang-sqlite3 versions got messed up, causing ejabberd build to regress
to 1.1.6 from Jan 2018 when using rebar3 which fails to build with OTP23.
Update to correctly tagged version 1.1.9, which also has matching hex.pm
package, fixes this.
