Change-Id: I54aae60a5ad987b7a9fd74c012f337952b855ec5
Reviewed-by: Ulf Hermann <ulf.hermann@qt.io>

Change-Id: I294cb27a59754f7090f9bc1a500faf0423571549

We should really backport this to 5.14. Someone will hit it.

(cherry-picked from commit 78fd438f158839ffebcd52cc7974eac28489dbdd)
Change-Id: I2c713fd759ac40aaaac0c0943edb993d3e27686b
Reviewed-by: Fabian Kosmale <fabian.kosmale@qt.io>

- the test items must be visible, so that waitForRendering() works
- arcTo::test_paint tried to render an out-of-bounds arc, which
  resulted in rendering nothing.  Now renders within the 100x100 canvas.
- painted() is not emitted the first time the Canvas is rendered.
- Canvas.save() saves relative to the directory from which the test is
  run, while Canvas.loadImage() loads relative to the test data
  directory in this autotest (other tests are loading red.png for
  example).  So we need to use absolute paths to test loading and saving
  in the directory where the executable is.
- canvas.getContext('2d').getImageData(8.5, 8.5, 8.5, 8.5)
  now triggers different rounding behavior in QRectF::toRect(),
  after qtbase 88e56d0932a3615231adf40d5ae033e742d72c33:
  it becomes QRect(9,9 8x8).  The assert in qt_create_image_data()
  needs to accommodate that.
- Fixed another pedantic warning in qt_create_image_data a few
  lines above: if it creates the image itself, it needs to round
  the qreal width and height values.

This reverts commit a23ee5c0de0d91859e1e76e64073861347dd9861 and
amends 424cfef3cc3c140df51905713fa3849562bc494d and
d142b2d212ea09a7919a0a2761ee9c04d5c9bda8.

Task-number: QTBUG-41043
Change-Id: I825c2c5a2bbc8d5324c3ba41a681aa68bc25a159
Reviewed-by: Shawn Rutledge <shawn.rutledge@qt.io>

Task-number: QTBUG-41043
Change-Id: I3a48439d30d9ec1cd908197c8d63984c95d336e3
Reviewed-by: Shawn Rutledge <shawn.rutledge@qt.io>

Since the commit a18ab2a3822e0d, we promote the [startItem] in focus
tab chain when it is invisible to prevent endless loop. However the
problem still happen if the [startItem] is equal to [firstFromItem]

Fixes it by compare the [current] item with the original start item

Fixes: QTBUG-81510
Change-Id: Iae0207f39e2b8c4fc6ed0cf36f0a855668accfba
Reviewed-by: Mitch Curtis <mitch.curtis@qt.io>
Reviewed-by: Liang Qi <liang.qi@qt.io>
Reviewed-by: Shawn Rutledge <shawn.rutledge@qt.io>

Some fonts will still render something for characters below 0x20 so
these should not be replaced with a space and should just be left as is.
It will not be rendered as a box if the dontPrint flag is set so we can
trust the renderer to do the right thing here.

Change-Id: Ie091c14713d8c2948c82b1991c295d80bc35dfde
Reviewed-by: Eskil Abrahamsen Blomfeldt <eskil.abrahamsen-blomfeldt@qt.io>

Since the type for a TableModelColumn can be seen as an int when it is
in the first row, it is still possible that it is representing a double
value in other rows. Therefore it should be allowed to add/change a row
that would display a double for that column.

Change-Id: I994a8ead595f836a20a4e82fbf94953b1aa4b7da
Reviewed-by: Mitch Curtis <mitch.curtis@qt.io>

In the dtor we don't need to care about any side effects a direct
delete may have. Rather, any deleteLater() may not take effect
anymore as the event loop may be gone already.

Task-number: QTBUG-82000
Change-Id: I97935dc47fbbfd0c050e80c333c36a05f685c45d
Reviewed-by: Joni Poikelin <joni.poikelin@qt.io>
Reviewed-by: Simon Hausmann <simon.hausmann@qt.io>
Reviewed-by: Ulf Hermann <ulf.hermann@qt.io>

The reusable items pool should only hold unreferenced objects.
Therefore, we can immediately delete them when draining. release() is
not suitable here because it unconditionally decreases and therefore
underflows the refcount.

Furthermore, the metatype is also refcounted, which means we should
keep it in a QQmlRefCounter in order to not leak references.

Task-number: QTBUG-82000
Change-Id: Iefdaaecc34342eb2e3b1e5a3281f2e46ac472347
Reviewed-by: Joni Poikelin <joni.poikelin@qt.io>
Reviewed-by: Simon Hausmann <simon.hausmann@qt.io>