Looking at when Invoke was called from Idle (or from auxiliary Timeout
if there was no chance to run the Idle within 100ms) and when duplicate
merging and modification of the queue is done before getting dispatched
and cleared; Then the pattern is that duplicates are collected and
merged during the current block of events to be processed, and
dispatching the flush via PostUserEvent instead gives the same results.
During startup, scheduling via PostUserEvent drops the need to have the
aux timeout, and while the number of messages seen in the first queue
flush (now via PostUserEvent instead of 100ms Timer) remain the same as
before, subsequent queue flushes can get processed earlier while
smaller, though typically at the same time as before once the document
is fully loaded.
Change-Id: I64c6bbc3dea9fb3a512c2a9521303a8f143d4a89
Reviewed-on: https://gerrit.libreoffice.org/c/core/+/167064
Reviewed-by: Michael Meeks <michael.meeks@collabora.com>
Tested-by: Jenkins CollaboraOffice <jenkinscollaboraoffice@gmail.com>
Reviewed-on: https://gerrit.libreoffice.org/c/core/+/167307
Tested-by: Jenkins
Reviewed-by: Caolán McNamara <caolan.mcnamara@collabora.com>
e22e7ca948