Change-Id: I774864241bc15fcfd08cd2becbf34aae0349a9d8

"perf stat" says:
Before  14,009,233,975,201      cycles
After   13,660,876,595,941      cycles
i.e. shaved roughly 2% of the cycles

Change-Id: If604a125a8a5040281abd699678d0c791d5bbb51
Reviewed-on: https://gerrit.libreoffice.org/20350Reviewed-by: Noel Grandin <noelgrandin@gmail.com>
Tested-by: Noel Grandin <noelgrandin@gmail.com>

This reverts commit 103cd686.
Better fix for tdf#96012 in b2d47985
("simple mode" for the multiselection)

Change-Id: Ia6d7e5132e3f1997906a03f67ea4d7d0cf54ce7a

Change-Id: Ia4d4c56318bc7d32d317abbda27eda9d3413ba75

Disable multiselect and don't manually select the entry

Change-Id: Id04e757e4d2031718f95af1d0f512b4ca2df205c
Reviewed-on: https://gerrit.libreoffice.org/20360Tested-by: Jenkins <ci@libreoffice.org>
Reviewed-by: Julien Nabet <serval2412@yahoo.fr>

Change-Id: I2fdbc2ac10f483eee154bdf69479ba217a91ef7f
Reviewed-on: https://gerrit.libreoffice.org/19605Tested-by: Jenkins <ci@libreoffice.org>
Reviewed-by: Michael Meeks <michael.meeks@collabora.com>
Tested-by: Michael Meeks <michael.meeks@collabora.com>

A Writer document can have multiple arbitrary RDF graphs, handle the following
subset:

- instead of handling all graphs, have a whitelist, look at the TSCP graph the
  DOCX importer creates
- look for (paragraph, key, value) statements and export those

Change-Id: I142c24fb4c8140bde5fc1efad2932886de167b90

The purpose of this class is to provide access to the subset of the RDF
metadata that's interesting for core code and for internal filters.

Change-Id: Ibecba302dd839b537a36b9f7a15f012c6ea26869

This reverts commit be0bf120, which is no longer
needed after 12fc5b85 "unnecessary include
brings in boost headers that are broken under rhel-7"

Change-Id: I1ad477b9db9ba960ae2c417799e5d91b63538d57

At the same time replace old OSL_TRACE, OSL_FAIL, OSL_ENSURE
with SAL_* as appropriate:

OSL_TRACE =--> SAL_LOG or =--> SAL_WARN if requires some kind of
attention by the developer.

OSL_FAIL =--> SAL_WARN or =--> SAL_INFO if the warning is brought up
later in the program flow.

OSL_ENSURE =--> assert oe SAL_WARN_IF when appropriate.

Change-Id: I1cf8f76acdec6f37746488b22cbf579802a9d79d
Reviewed-on: https://gerrit.libreoffice.org/20034Tested-by: Jenkins <ci@libreoffice.org>
Reviewed-by: Stephan Bergmann <sbergman@redhat.com>

Patch by: me

(cherry picked from commit d46d79f2)

Change-Id: Ibd175c65babdde48132692fd1979a00929356bb4

Change-Id: I58427aafb30eca64a67bcf7dd9a7738d8da3e085

Change-Id: I8c8589f6ff7a190556cc4aa016ae343a19378b18

that way the "preferred" image format for a platform
for icons etc remains sticky

Change-Id: Ia76af0b7c4f27650038cab214b6406b8ef1a5fc4

Change-Id: Id2ed21b397a3f56413c344dcf9211ab64a939286

Change-Id: I4f07c7c61d324a85f9f5dee0f25a7358facebd41

Change-Id: I9ed8586e8b77b009d55e411fdaa863eefc38b1c2

This reverts commit 7e9bdfe6.

Seems to be causing problems for people using 'ssh -X'.
I suspect that it has something to do with PaintHelper's destructor.

These can be sort of arbitrary key-value pairs around one or multiple
runs, handle the following subset:

- when they appear at a paragraph context -> we assume they are
  annotations on the paragraph
- when the smart tag's URI/element is RDF -> we map these to RDF
  metadata statements on paragraphs
- when the attribute name's namespace is known, because in ODF we need
  to specify both a path and a type (namespace) for the RDF graph, and
  OOXML only provides a namespace

As a start, recognize the TSCP BAF namespace from
<https://www.tscp.org/wp-content/uploads/2013/08/TSCP_BAFv1.pdf>.

Change-Id: Ib188b1395e7ec7e0441b4f12f86cfef99fb9f096

Implementation was removed in 630ec711
(tdf#92925 Remove XPM export, 2015-12-01), this would only fail at
runtime on the desktop when doing a dlopen(), but gets detected on
Android at link-time.

Change-Id: I0345c23ef52fb87c9d56bc54490ba0302fc08ac1

transition would be finished for breeze

Change-Id: Id159b09f7d62ca087a946e19b5568424bfe9ca0c
Reviewed-on: https://gerrit.libreoffice.org/20345Tested-by: Jenkins <ci@libreoffice.org>
Reviewed-by: Michael Meeks <michael.meeks@collabora.com>

Commit c91feb2f (fix Android build,
2015-12-02) was the right direction, but probably was a blind fix, so do
the same at the remaining places as well.

Change-Id: I49637fbb9ab848de7c465724387a8676d2d98de0

Change-Id: I9d74642156d7bb79577671bae24346c75178c5c7

Change-Id: I79e065f0c68b149d2ef69f428d31e36e97a6098b

...since 5b4d4af3 "bin agg"

Change-Id: If86d294f812a670917b181f3bd140883ad2df71d
Reviewed-on: https://gerrit.libreoffice.org/20348Reviewed-by: Stephan Bergmann <sbergman@redhat.com>
Tested-by: Stephan Bergmann <sbergman@redhat.com>

Change-Id: I66b2ba5d8e419f6af5eb3f85f8c12effcdaf4d5e
Reviewed-on: https://gerrit.libreoffice.org/20319Tested-by: Jenkins <ci@libreoffice.org>
Reviewed-by: Noel Grandin <noelgrandin@gmail.com>

...when (non-negative) QueryOp enumerators LESS/GREATER_EQUAL are crammed into
a 2-bit meOp bitfield, where enums are implictly signed for MSVC, so the values
actually storable in the bitfield range from -2 to +1.

The clang-cl warning would go away when fixing the underlying type of QueryOp as
unsigned, but then GCC would start to emit "error:
‘ScLookupCache::QueryCriteria::meOp’ is too small to hold all values of ‘enum
ScLookupCache::QueryOp’ [-Werror]."

So don't bother with bitfields at all:  For QueryCritera, for one there's a
union member with a double and a pointer, so sizeof (QueryCriteria) will be
twice the size of double anyway; and for another, MSVC doesn't combine bitfields
of different type, so the bool members were separated from meOp anyway.  For
QueryKey the reason for a bitfield is even less clear cut, and it might only
have been there so that comparing (negative!) values read out of
QueryCritera::meOp compare equal to values read out of QueryKey::meOp under
MSVC.

Change-Id: I69fb068bea914c00a29001155218cb9f1b8f8a9a

...for now; workdir/UnpackedTarball/icu/source/tools/toolutil/Makefile invokes
the compiler with a -DU_HOST=\"...\" argument, and apparently directly executes
the compiler from CreateProcess, not going via a shell invocation for the recipe
line.  This confuses clang-cl for whatever reason, and for whatever other
reason, forcing make to go via a shell invocation (by adding "true &&" into the
recipe line) fixes it.

Change-Id: I3757a8856f93228c19475b37f3037fa9519a426f

* Avoid errors about (somewhat different, but apparently in a way that MSVC
  doesn't care) redeclarations of system header stuff (and working around that
  with BOOST_USE_WINDOWS_H would lead to other problems).

* Need the workaround for broken MSVC 2010 STL iterator taxonomy for clang-cl,
  too.

Change-Id: I4f9c9b5cc8a19856cc409450aac1fb1baa5554c9

Change-Id: I819be5efb25632d26fe49a71dbc07fe16e4914b1

Change-Id: I23da54974f39da5fccb619d6fa68eff38e70f5a5

Change-Id: I0d24bcdfeb0d004607569da089c9f787a868da72

Change-Id: I596dc0e2306e7170e65c77a1635f88407272e6b1

Change-Id: I8e75b0ab2439592316fc0d871280a438e3ae2f1c

...like CreateMutexA redeclared in
workdir/UnpackedTarball/boost\boost/thread/win32/thread_primitives.hpp.  As such
problematic Boost include files are apparently not included from LO-proper
source files, only from external, it looks easiest to just silence that with
BOOST_USE_WINDOWS_H here.

Change-Id: Ia5ec2325934e6d7fdcf91e6faa2e671aee2091ae

Change-Id: I068ff0ef2252409689c0c376ec41bdd97b4567cc

...included via external/poppler/poppler-snprintf.patch.1

Change-Id: Ibe20452e7da4995ca33c97dd99dc441060e73894

...with the executable specified in the first arg, instead of as part of the
second arg (i.e., the command line)

Change-Id: Ie6e232f6880b5bfbb91a52ee5398b91a0ccddc4d