Cherry picked from 2674376d

Change-Id: I882f6b516f617f37f2b966cb2e8c78718895e305

cherry-picked from 24b6225f

Change-Id: Ic0624618591285544e3fe4032ba9ac12a96ee003

Cherry picked from 405454c8

Change-Id: I18bb06ff9dfffcdb2dd958c736b6e81af0b2b4ca

cherry-picked from 95357d31

Change-Id: I1d403859091a6e59b1045126c5db2bf2ffd573e1

cherry-picked from 969d60a4

Change-Id: I074f2f024624aafea4456304295d747318755ffd

cherry-pick from da5f1268

Change-Id: I1e462668270f33f7f7f2082b20c12c18d0ad8042

Change-Id: I013893c7ca4b9d4364d2d19fdd0440c7dcf08910

Change-Id: I8105fc255e6ad7d9d102cbd1928329453b9ca7e8

Added dialog for smooth (gaussian blur) filter which now accepts
a parameter for setting the strenth of smoothing (bluring).

Change-Id: Ida6709b060cb5429a63af1994493e716fd0bfebb

Change-Id: I4ad84a293f662a6e191758c30660ac11142aa3eb

Currently implemented Gaussian Blur filter uses a static matrix
which means that the filter can not be parameterized. The new
implementation of Gaussian Blur filter and accepts a blur radius
as a parameter so the user can change the strength of the blur.
For this the blur matrix is generated at each call. The new Blur
implementation reuses separable convolution from Lanzcos rescale.

For negative values of radius the Bitmap will be sharpened. For
this an Unsharpen Mask filter is used, which is actually a blurred
image substracted from the original image.

Convolution calculation now uses 2x transposing horizontal pass
instead of horizontal and vertical which simplifies the code by
reducing code duplication.

Instead of more convolution methods, only one generic method is
used and kernels are now a parameter to the method. For this
Kernel class was introduced with responsibility to calculate
kernel walues and hold kernel specific information.

Besides Lanzcos resampling also Box, Bilinear and Bicubic resampling
was introduced. Box is the simplest and fastest but with still
good quality.

Change-Id: I41a11bda977cbd5b4b2352776f06fba0e0457a43

Change-Id: If91c33dad811d5c9a34581773f7f6a7f86dd82bd

We were ignoring redlined (deleted) text when counting characters, but not when counting
words or characters without spaces. We now mask all redlined/hidden content from the
expanded node text before operating on it.

Change-Id: If882a6dde883a2ba0ca99909ca7ac2b9abd4a1fb

Change-Id: I3358fad4949c2c21ecf63983a36b7c8428df9f06

...which also needs a change to the logic that symlinks *.jnilib to *.dylib, as
that would have symlinked pyuno.so onto itself (as it contains no .dylib).  (And
while we are at that, anyway, also only store relative paths in the symlinks.)

Change-Id: I6f3e9effc4d185df935795958cc84e60e862a424

Change-Id: I90cba97c112af81ca50feca5bfa70d0d651b6083

Change-Id: I88de7a942fbc9e0c51a3261236f5203f037d2392

Change-Id: Ic727e2a6a30c16b3467b8c169512a7cb2757f146

Change-Id: I386725e07c34336249dfc935b1e9f9624d1d3be8

Change-Id: I611dc412b5461f368f23aa8239e48d0e175f5168

Change-Id: Ie269fb800390dbc3f37f0002509ebcccafb038de

There are still some pch headers in use. Simply:

find . -name *_pch*

Change-Id: Idafcd4f1451877ad9b08c8bbea526266d7414553

Change-Id: I043e226a1cec527c51cbd91075634d5c9ca371d8

Change-Id: Ib763b8ab1e728d400859a7abb355bba182b28684

Change-Id: I172073bac10e8d0449c0f5c0a226dd7ace013376

Change-Id: I137c0a42ab8413213d93f1bbd868077e2c042ce0

Since Excel doesn't support named database ranges, I also added tests for
unnamed database ranges for ods, xls, and xlsx.

Change-Id: I491324883c95cf2edb13c0561e975a13d13ba7d7