| Age | Commit message (Collapse) | Author | Files | Lines |
|---|
|
Before telepathy-glib 0.20.3 and 0.21.1, we had this incorrect sequence
(pseudocode) for each group:
* NewChannels(the group)
* GroupsChanged([the group], added: [...], removed: [])
* NewChannels(the group)
In 0.20.3 and 0.20.1, we removed the second emission of NewChannels.
Unfortunately, that broke this test, which was specifically expecting
GroupsChanged followed by NewChannels.
Rather than reversing the assumption, I'm doing it properly, by
expecting the events in no particular order.
Signed-off-by: Simon McVittie <simon.mcvittie@collabora.co.uk>
Bug: https://bugs.freedesktop.org/show_bug.cgi?id=67828
Reviewed-by: Guillaume Desmottes <guillaume.desmottes@collabora.co.uk>
|
|
|
|
I previously believed that I had added code to WockyPorter such that, if
someone sent us an IQ with no id='' attribute, it would be silently
dropped rather than being handed up to the application. (This would
technically be valid behaviour; RFC3920 §9.2.3 “IQ Semantics” says that
“The 'id' attribute is REQUIRED for IQ stanzas.”
<http://xmpp.org/rfcs/rfc3920.html#stanzas-semantics-iq>)
I mentioned this to Rob, who was horrified, because Gabble's roster code
specifically handles roster pushes lacking an id='' attribute because
this happens in the wild. I checked, and I had apparently imagined that
I had made the above change to WockyPorter.
This patch adds a regression test to check that Gabble correctly
processes roster pushes without an id='' attribute, and doesn't crash
in the process.
|
|
This changes the signals we get slightly, because TpBaseContactList has
more context.
|
|
send_roster_push)
|
|
|
