Where "unexpected" means the host in 'Host' header is not
a registered route. The rationale is to avoid propagation
of uknown "Host" further in the code, which may lead to
nasty errors related to reading configuration values, calling
functions from ejabberd_router.erl, etc.
Now every such string MUST be encapsulated into ?T() macro.
The macro itself is defined in include/translate.hrl.
Example:
-module(foo).
-export([bar/1]).
-include("translate.hrl").
bar(Lang) ->
translate:translate(Lang, ?T("baz")).
Don't just adhere to the node's access_model when delivering the last
PEP item but also when a new item is published.
Many thanks to Philipp Hörist for spotting this.
Option 'captcha_host' is now deprecated in favor of 'captcha_url'.
However, it's not replaced automatically at startup, i.e. both options
are supported with 'captcha_url' being the preferred one.
mod_privacy updates the c2s state in user_receive_packet, which
tracks the *result* of the IQ set for active privacy lists.
When a second stanza is sent directly after a privacy list request,
the second stanza will be processed using the old privacy list,
because the IQ result has not yet been routed.
The value of monotonic_time/1 may differ significantly between
nodes in a cluster due to different erlang:time_offset/0 values.
Thanks to Christoph Scholz for spotting this
Since commit 633b68db11,
ejabberd_auth:try_register/3 returns {error, exists} rather than
{atomic, exists}.
Thanks to Thomas Leister for reporting the issue.
Revert the changes applied to mod_stream_mgmt in commit
b76f90fe39, as the new implementation of
mod_offline's 'use_mam_for_storage' feature doesn't need them.
Previous version was trying to determine range of messages that should
be fetched from mam by storing time when last user resource disconnected.
But that had couple edge cases that could cause problems, for example in
case of node crash we could not store information about user disconnect
and with that we didn't have data to initiate mam query.
New version don't track user disconnects, but simply ensure that we have
timestamp of first message that is gonna be put in storage, after some
measurements cost of that check with caching on top is not that costly,
and as much more robust i decided to introduce that change.
