GitLab

This needs to be fixed in Akonadi, as we call clear() there.

This reverts commit 159abaf2.

In my data, most often the final size of fetchResponse is 16.
By reserving that data upfront, we can get rid of 3/4 of the
list allocations, according to the growth-strategy outlined here:
http://qt-project.org/doc/qt-4.8/containers.html#growth-strategies
I.e. before, we'd allocate 4, 8, 12, 16. Now we directly allocate
room for 16.

REVIEW: 121408

In EntityMimeTypeFilterModel::filterAcceptsRow, we can check the
mimetype first, before checking whether the item has a payload.
The latter requires a temporary data allocation inside QVariant
for the Akonadi::Item, whereas the former can leverage the optimized
QVariant code for QString, which is faster.

REVIEW: 121407

This data is static, and should be declared such. Doing so gets
rid of thousands of temporary allocations.

REVIEW: 121405

This showed up in heaptrack, but was actually due to another issue
elsewhere. I still think this optimization does not hurt. But if
you don't want this, it could be skipped, since the following
optimizations will remove this as a hotspot anyways.

REVIEW: 121404

The job never emitted finished() when the CollectionFetchJob returned no
collections. This causes KMail to get stuck when user opens an invitation
without having any calendar resource set up.

BUG: 297930
FIXED-IN: 4.14.4

Reset cut action when we call another cut action. Otherwise we can set "Cut" action on all item which will not cut

Now first runs don't cause KMail emergencyExits when special collections
are created before there are resources created.

During loop event a signal refresh password.

Reviewed-By: Dan Vrátil

Otherwise we potentially use outdated attributes during the sync.

Reviewed-By: Dan Vrátil

Otherwise setEnabled(true) will still not result in the value getting
updated (since the default value is true as well).

Isolated unit tests that are testing moving items in Akonadi would get weird
change notifications (MOVE, but the item had different RID, and then ADD of
the original item into the original collection, because Knut did not remove
it).

Surprisingly no Akonadi unit tests seem to hit this problem (probably not
tests are checking other Monitor singals and are only comparing IDs, not
RIDs), but Zanshin guys ran into this with their tests.

It was only used by the QML GUI, and it triggers a large amount of dataChanged
signals in a model that has many many layers of proxymodels on top.

Federal elections in the United States are held the first Tuesday
after the first Monday in November. For the purposes of this file,
that means the Tuesday on or after 11/2.

REVIEW:120440

BUG: 337292

This code is called very often while loading a folder (with threading enabled)

Maybe in frameworks these methods (mimeTypesMonitored etc.) should return QSet
to avoid the conversion altogether?

REVIEW: 120552

The ItemSync must never emit result before it's complete. This patch fixes
this behvaiour in case of an error on the ItemCreateJob.

* Adding a job does not reset the status if a rollback was initiated already,
and thus avoids a second rollback on commit() after one was already performed
earlier (this resulted in the "No transaction in progress" error message).
* The rollback job, that is equally added using addSubjob still gets regularly
executed.

REVIEW: 120527