Apparently some servers don't mention "EXPUNGE" in untagged EXPUNGE
response. This wasn't handled correctly due to a typo in response
length check. The code no longer crashes in such case, but will now
print "Unhandled response" warning, which is OK, since we don't
handle results of the EXPUNGE command anyway.

BUG: 364045
FIXED-IN: 16.04.3

The else branch would get evaluated when OK STATUS response
is parsed and it wrongly assumed that to be an error situation,
while that's a completely valid case.

Secondly we should not assert on (unexpected) responses from the
network, as any kinds of error might occur.

I'm wondering how nobody else is hitting this assert in production
and we are only seeing this on the CI.

ignoreSslError() is async and the whole Session including SessionThread
can be deleted in the meantime. Checking against member variable of the
potentially deleted object does not make much sense since the memory can
already be overwritten.

Instead we hold the pointer in a local QPointer variable, which will get
updated when the SessionThread is destroyed while we are waiting for
ignoreSslError() to return.

BUG: 332830
FIXED-IN: 16.04

RFC 2971

As suggested in ECM documentation.

[-Wclazy-qgetenv]

REVIEW: 125426
CCBUG: 351850