GitLab

This fixes TESTRUNNER_DB_ENVIRONMENT not being set because the multiline
value for QT_LOGGING_RULES broke the sourcing of the shell script.

Reviewed live by Dan Vratil.

If the ETM's Monitor explictly monitors only certain mimetypes, then
only populate Collections matching the given mimetypes. Otherwise
we might end up populating collections that we know about only because
they are ancestors of collections that match the mimetype filter.

ServerManager automatically handles Akonadi instances and includes it in
the service name.

MSVC does not like when the same object files are reused for both
shared and static linking, so we have to compile the sources twice.
GCC/clang don't mind, so we keep using add_library(OBJECT) there.

Otherwise MSVC complains that the friend declaration has different
visibility from the original declaration.

The XSL transformation automatically includes the class name when
generating the code, so thi way we were ending up with names like
Part::Part::Storage instead of Part::Storage, which GCC seems to
handle, but MSVC does not like it.

When Akonadi is started each application and agent starts requesting
its Collections and statistics. This causes the CollectionStatistics
cache to query the stats for each Collection individually with the
very expensive query. On my system all the statistics queries together
take over 17 seconds.

With this change the CollectionStatistics cache simply pre-fetches
statistics for all Collections in single a query, which is MUCH
faster (less than a second on my system), because the expensive JOINs
only have to be done once.

As a result of this change applications and agents become responsive
much faster after start and CPU and IO load during start is massively
decreased.

... since the passed item.parentCollection() is already the target
collection

Differential Revision: https://phabricator.kde.org/D5711

(cherry picked from commit 33d86e16)

Reviewers: dvratil

Reviewed By: dvratil

Subscribers: #kde_pim

Tags: #kde_pim

Differential Revision: https://phabricator.kde.org/D6021

BUG: 379809
FIXED-IN: 5.5.2

CURRENT_TIMESTAMP is now represented as QDateTime::currentDateTimeUtc()
(see a9437332)

... since the passed item.parentCollection() is already the target
collection

Differential Revision: https://phabricator.kde.org/D5711

Differential Revision: https://phabricator.kde.org/D5672