Change-Id: I8f1f61110d06333bc1140caf13d9e96e669fc231

Change-Id: I0c3a565a3ae276fb60bf99ec32941715f1556127

Change-Id: Id094ffa7972df0477ad26b2fc579fe010e745550

Change-Id: I8c1331c32baf92b077f45678929df0e4cf185774

No markers yet, but we're aware of the selection rectangle at least.

Change-Id: I92d3fc62d066e401ad19107fbbd414b1b9b26125

Change-Id: I4dbee4048f52f480efef4454d3714862e4ddcf4b

The problem was that the earlier "SwEditWin: disable map mode in
LogicMouseButtonUp/Down" commit disabled map mode only during the
handling of the mouse event, so in case a method invoked by the idle
timer used our logic coordinates later, it still performed the pixel ->
logic conversion, leading to jumping cursor after a mouse click.

Fix this by disabling map mode early, even outside the mouse handlers.

Change-Id: Ib9f194f5555536a9af21a2d66bd7345addee72c3

Change-Id: I36c359b930a09dbcf51880d9488f335cc38fba45

Change-Id: I294e1d8382d997b70cb4e5af13cdef5d606885ae

Change-Id: I42c0971b10e3926f8fca989884da089bec8d59d4

Change-Id: I78c8a984907d1481134426eef164f2587c0f06c6

On one hand, this cleans up duplication, but what's more important is
that without this, svx can't invoke sw's LOK callback, which will be
needed for graphic selection.

Change-Id: Id414489fc3fbc52914b28fdb9bea4134edfbeafe

The only user of SwCrsrShell::SelectCell() can actually use SelTblBox()
instead.

Change-Id: I286a3d8c25834dd08897f0cfc37a3ec31a89ce7a

With this, if a Writer document contains an image, then clicking on the
image then back to the text properly shows/hides the blinking cursor.

Change-Id: I3313319ded614a2eb86e1c66c926189511a7cca6

There are multiple levels of being hidden: if the cursor is not visible
due to just blinking or in case it won't show up after some timeout,
either.

Let's use CursorOverlayVisible for the condition if the cursor should be
painted right now or not, and CursorVisible for the long-term "is it
visible" state.

Change-Id: Ib583236c1429f25e14ba79fdabad4ea899bb4269

This partially reverts commit 3d4c1b9d
(Initial SwEditWin::LogicMouseButtonDown/Up, 2015-01-21), which tried to
catch all pixel -> logic conversions in SwEditWin::MouseButtonDown().
The problem is that it doesn't really scale, we can't do the same for
drawinglayer as well.

Instead, simply disable the map mode in the Logic...() methods as we
already do in Impress, which doesn't have a single central
LogicMouseButtonDown/Up method, so there we had to use mapmode disabling
from day 0 for tiled editing.

Change-Id: I72be546400f5db10ac6d09a9bb10f7f0658942a5

The main problem was that UserInstallation did not point to the usual
workdir/unittest/, so initialization failed early, not finding the
soffice.cfg directory under
workdir/CppunitTest/libreofficekit_tiledrendering.test.user/.

Disable one testcase that still fails, though.

Change-Id: I3fdae5ead47dd2a805a0ed8d3cb407938f868e3b

Cppunit tests just don't run the desktop logic to determine whether to
restart normally, but for LOK, that happens.  Without this, if versionrc
changes, a LOK-using unit test would fail.  Simple reproducer:

sed -i "s|\(buildid=\).*|\1$(echo $RANDOM|sha1sum -|sed 's/ .*//')|" instdir/program/versionrc
make -sr CppunitTest_libreofficekit_tiledrendering

Should be an acceptable alternative approach to commit
5acadbfb (Implement lokconf_init.,
2014-07-21).

Change-Id: I4354e646d3bcc0c1199db5a175708206b4a30294

A double click is a set of 4 events, only one of them was sent, that's
how it worked in gtktiledviewer, but not on Android.

Change-Id: Id0eaef3765d385aab588b834f66b9f4b4f3ccb70

Change-Id: I40df42b941ab1fbf9d847772e8b5c0ddea5cfcb7

On the desktop, the logic is that if we double-click before a word, then we
select the word, but if we double-click after it, then we don't.

When tiled rendering, it's more important to always select something, as
that's how the selection can be adjusted further. So if we long push
after a non-empty text node, then select its last word. This way
something is always selected if e.g. the user long pushes inside a table
cell. (If the table cell is empty, then that's handled in
SwCrsrShell::SelectCell().)

Change-Id: I6e84874b8cf252fc6b72a75ffe9db67b03636862

When not tiled rendering and the user double-clicks on a word, then
that's selected. If the double-click happens in an empty cell, nothing
happens.

When tiled rendering, still try to select a word, but if that doesn't
work, then see if we are in a table, and then select just that cell.
Then the user can use the usual selection handles to adjust the
selection further.

Change-Id: I6beabf5c73a02f1eb2ede6f8cf9062c1e6610e9e

The shell (i.e. non-table) cursor does this already and this also fixes
the SwRect::SVRect() assertions firing during the drag of a table
selection handle.

Change-Id: Ia63844f5e2f6e107d47b7ed39af2a72bd433003e

Text handles are usually shown under the current selection or
cursor position. Previously this was done in InvalidationHandler,
now the TextSelectionHandle takes this into account just before
drawing. The reason is that the repositioning needs to be taken
into account when the handle moves to reposition the coordinates
back to original position.
In addition to that, the number of times the move handle update
event is send is now limited to 50ms to reduce stress for the
device (less in queue, less messages via LOKit). This also makes
the handle more fluid and jump around less.

Change-Id: Idc401375c82b3eeac66328d98c80bf3cb7b675ea

Change-Id: Ic60463027c7f1b2d513539adba834b8d8e0ce6d3

Change-Id: I04def21b684d0c95bc3a26250fc8af42bc1b1fa6

Change-Id: Ia360b2461d0bf31abc62ae443822259b7197d90d

stopgap measure until tdf#57873 is fully implemented,
since these days by far more people use Thunderbird
than Seamonkey (or Firefox address book, if the latter
even has any...).

Change-Id: I638ec90277063b1e7536c045bffb1052d3b76911
Reviewed-on: https://gerrit.libreoffice.org/14793Reviewed-by: Lionel Elie Mamane <lionel@mamane.lu>
Tested-by: Lionel Elie Mamane <lionel@mamane.lu>

Change-Id: Iee32ede1cc474a361fb3b1bf47f1c52b4ff2c4b7
Reviewed-on: https://gerrit.libreoffice.org/14799Reviewed-by: Miklos Vajna <vmiklos@collabora.co.uk>
Tested-by: Jenkins <ci@libreoffice.org>

Change-Id: I921ffaf2ee4be73ecfffca8bc36cfae484b70a32

Change-Id: Ibc378fa5f515de61bb768b4ef082638b40c94e00

Change-Id: I883c0b95d91ae9cb7e59b8f68ae0c8a3ebcb9f88
Reviewed-on: https://gerrit.libreoffice.org/14796Reviewed-by: Miklos Vajna <vmiklos@collabora.co.uk>
Tested-by: Jenkins <ci@libreoffice.org>

Header on second page was missing because document was parsed as document with facing page.
Original document contains \facingp0 token toggling facing page off. However LibreOffice treats
it as a boolean switch. This absolutely conform RTF specification (1.9.1), but differs from actual
behavior of MS Word: actually this RTF token behaves like \b token (\b or \b1 are enabling bold text,
\b0 - disabling).

Change-Id: Iea2b52cde311ab1b02938461ba0945191339c783
Reviewed-on: https://gerrit.libreoffice.org/14743Tested-by: Jenkins <ci@libreoffice.org>
Reviewed-by: Miklos Vajna <vmiklos@collabora.co.uk>

Change-Id: Ia51cea076e5368956bcd87dd201d15eafbf6ff24

This reverts commit 79b3e125.

Change-Id: If5ab5dbf638d1ff81693fafc1895e68314929f90

Change-Id: I88bff880a5fb27b259b66984a0637fe7d2ab51f6

Project: help  00946ca3f8cd3837129c714e025d6ca2ae783ad8

Change-Id: Ia1d37d150c2186f014733cd87179445722b54346
Reviewed-on: https://gerrit.libreoffice.org/14789Tested-by: Jenkins <ci@libreoffice.org>
Reviewed-by: Lionel Elie Mamane <lionel@mamane.lu>

Change-Id: I8fa5cc15e2d69b572f59f254eeeae051865a5b5f
Reviewed-on: https://gerrit.libreoffice.org/14785Reviewed-by: Miklos Vajna <vmiklos@collabora.co.uk>
Tested-by: Jenkins <ci@libreoffice.org>