Change-Id: Ifaeb87427d6e2e0c2bb0fcd19e0d39bf15c76973

editeng sends LOK_CALLBACK_CURSOR_VISIBLE after the text selection is
already created. The effect of this is that after a long push on a word
you get not only the selection overlay + the start/end handles, but also
the middle one, which is not expected. Just check if we're in a
selection already, and if so, don't react to
LOK_CALLBACK_CURSOR_VISIBLE=true.

Change-Id: I939c957feb7889dfb1997fac1051cb4eaa77004c

Also, if we are at it:

- clean up 'run' as well: since the doc browser is the default activity,
  no need to pass the test doc path anymore
- make 'install' not depend on build: a full build would need a toplevel
  'make' anyway

Change-Id: Ia55d52f767ab3e0be02a753a95b2aac02f8491cc

Without this, graphic selection was still around during text editing, so
long push on an editeng word (in a Writer shape) triggered the shape
movement action, not a word selection in editeng.

Change-Id: I2ec2f1f9a417bacd08d8b357b4203d48ebaf15c7

Change-Id: Iaecedd01d7e7bba90e49ddaa5f36859d9218242d

... and just error out with < 16G and all symbols. Should help newcomers
who have e.g. 8GB of memory, build with dbgutil, but then unable to
complete the build as the linker never finishes.

Change-Id: I9f3ad5c6d76a18d6aa5b0b2d61aa115ed441b8e3

When the bitmap with handles is not found (usually happens when
using LO as a library (for example on Android)), the returned
bitmap was empty. This caused that the hit test for the handle
always returned false. This commit replaces the empty bitmap with
a black 13x13 bitmap which should respond to hit test. This should
not influence normal behaviour when the bitmap with handles is
present.

Change-Id: I39330b1867abfb540c215f03b77e5eac2447f0dc

Change-Id: I985b7b22c4794862a421ad797b907e33266c2c27

With this, at least a simple .odg document is now loaded again without
crashing.

Change-Id: I47200ffec680aefc0ee7769f9311a550c0329dc7

Add constructors which are no longer part of sd_component_getFactory()
as of commit 4dafecfe (sd: convert
sd.component to constructor usage, 2015-02-28).

Change-Id: I209fdd9bc8c8b81cb7e69796a2afb248df856fc9

Based on a patch by Henry Castro.

Change-Id: Ic463c97577d521e3c2e4d6bc904c517aa564edca

Change-Id: I55e4eb784694de889d7bb0a3a6043a5a2b74239c

Change-Id: I119940f89284a19594819fb26f93155a476f487c

Change-Id: Ibf9f98968fd9007be6dc5481d1652923ecb98b89

It's not necessary since the normal widget renders normal 256x256px
tiles, not just a single huge one.

Change-Id: Ia1c6a6c8692769093232d66aa63fde096390a7b2

Change-Id: Ica316ed9660e235e756ef380da8f441c976de07f

Change-Id: Ic466e9efc0cb97fb39e94624f018ab5b98a54929

Change-Id: I49947cccfd2662533da3087cb1f04dc91014cfb0

Initiali implementation moving of graphic selection in
TextCursorView. Transmit start and end coordinates of the move
using a touch event with a special type "GraphicSelectionStart"
and "GraphicSelectionEnd" set.

In LOKitThread process the event and delegate the request to the
TileProvider (setGraphicSelectionStart/setGraphicSelectionEnd).

Change-Id: Iccceeb3d4de2fc1761b23fc5ede99d00d5e803aa

Change-Id: I26c3eb03ad3bcc23bed380e873d353bc2941de9a

Change-Id: I455e58f4395cb463efa888fe69ee7ea48b4681fb

This affects only the text area of the shape, single click on the border
still just selects the shape.

Change-Id: Ia3564e6a95213773512e4c904843ff42aeb3a9ba

Change-Id: I435bb683bae1634f156b3912e6120d479a917b00

With this, it's possible to drag the handle of not only the normal
cursor, but also the start/end handle of a range selection in editeng
text, e.g. Writer shape text.

Change-Id: I5b9f548fb6365ecb499c68cef0d89f7f8ff3a11d

With this, it's possible to drag the middle selection handle of editeng
text (e.g. Writer shape text) and the cursor position will be updated
accordingly.

Change-Id: I2b03c8543efca477f2e377f3bb93732cd1775f81

Change-Id: Ib1d4746d222c0a59a2ace0a97ce0fb1bac18f4ef

Change-Id: I46119fbae611a762930db9cccdd11c35788c9f9c

Change-Id: I992773c84eaf2dff63814e6223323fe99ae5a415

That was something from the other LibreOffice4Android that got merged
with LOAndroid3. The document browser doesn't use it, so we don't need
it anymore.

Change-Id: I5446eb7a0fe6e0de81532e981a0853da16910f54

Change-Id: I49d9bd7c702c260ea66ef211edbaf5e106264bb6

Change-Id: I8cb072b81e005e435577b617204e02fa94ff9f88

The problem was that when not tiled rendering, the selection is painted
with the inverse color, and so killing the selection can be achieved
with just painting it again -- using the inverse color again.

For the tiled rendering case, use IsInSelectionMode() to see if we're
killing the selection, and if so, just emit the empty selection.

Change-Id: I045f5b100ae8d4f5802b16f68376aa21de08b994

This order seems to be less confusing.

Change-Id: I95a7e3e2f6244915c820bb86e67745a777714e2d

Change-Id: I5ea37875aa36cc30847643721b369bbe491101e8

When we are entering some text into the sheet, now it happens at the correct
place.  Unfortunately we are still invalidating the wrong area, so it is not
really visible unless you tweak the ScGridWindow::LogicInvalidate() to
invalidate much more tiles.

Change-Id: I74a568761c3b21522fe291649c33829aaea200e7

Change-Id: I66b2fc7ee846d6865b11cf1cbb168a6e5fab2eea

Change-Id: Ib143870ff6e5cf9dd62b420d8f8f1e71dbac42e7

Also, add one missing 'break', which however did not have any effect on code
semantics, as the next case label only contained a 'break' anyway. (So I guess
it could be claimed that it is also a case of intentional fall-through...)

Thanks to fedor_qd for pointing these out.

Change-Id: I430a71b59c78391e92da7ca14290a4e3cdade496

Change-Id: Ic1a71851a80ed2715969b0f00a0e59ab3a0593db

Change-Id: I4bbfbd526f78c4806d6fc85a4011cbc4ef3a25fa