Change-Id: Iac01fc756a74cbaeebd1d9a01ddea74bbaf18f4b

Change-Id: I3f2781e2a041845e732046ea591a15c9bb21f1a5

Change-Id: I894fc64d6ae6881347bd1379d2934654536a01cf

Unfortunately requires nasm - the netwide assembler -
http://www.nasm.us/

Upstream libjpeg-turbo is meant to be built with CMake on Windows but
thanks to our gcc-wrappers we are able to avoid that.

jpeg is kept mostly for platforms we are cross-compiling to. For now,
it's used also for Mac OS X because jpeg-turbo does not build there,
which should be fixed later.

Change-Id: Id87b7072a8acc2578c3abf7e82cb1499e5094dbf

Another typedef that destroys code readibility. I've removed AnimViewList_impl
because it's just a vector of ImplAnimView pointers.

Change-Id: I97e5867e1a7aeae575e45989ba7436867da9c808

Severe overuse of typedefs makes baby Torvalds sad. :-)

AnimationBitmapList_impl is just a typedef to a vector of AnimationBitmap
pointers. It's used precisely once in the code - in animate.hxx. This just
makes reading the code more difficult - too many layers of abstraction.

Change-Id: If84892dc90220c7dfe8c071d629268efa226f069

Change-Id: Ifc7377fe1bb68b21e462818248706f53347b39b1

Change-Id: I82570a716a0dba6c466bc754ce66ac53e7c17efa

Change-Id: I55d6d1d403622e5ec04c29dd407274cf9db20c40

Change-Id: Idd2f223e0c83e0a6a19e70bd677058b6899bf766

just throw up hands in despair via std::abort instead of std::runtime_error

Change-Id: I356932ed2d76558bb4febbd2ac0a7bb6ea27bd0f

Change-Id: I1dbf7a73ec60115f11d3a5455de2871041cd1605

Change-Id: I1887a32aeef8097e3e8297d6bddc51179dd65127

Change-Id: Ic6a8f6435e00c7b679c58bdfe57da4555107787d

Change-Id: Ie778e541cecadeefc5ee5f30af9577a00c870a58

Sadly cannot forward declare "struct {...} TimeValue;".

rtl/(u)?string.hxx still include sal/log.hxx but removing osl/diagnose.h
was painful enough for now...

Change-Id: Id41e17f3870c4f24c53ce7b11f2c40a3d14d1f05

Change-Id: Ibbcdb4a3f46b7491396a1b1c7b4cdc1af05f30d3

Change-Id: I2eda8dab51e66af7856b55f2d5f297620e07ccdb

... to be used together --with-iwyu to track down bloat in headers,
which are not always named in the convention that include-what-you-use
expects for reporting issues.

Change-Id: Idffda7d49cca43170717004d52779b4663de19ae

Change-Id: I7b028ff3af6c5cd6c48b4cf6095a16b57745db8c

Change-Id: I5461a0d4e372260623ebb8833d60a9dc318757dd

Change-Id: Id4b01f02d5d90215164c07b8076498df65978fa3

Change-Id: I28b52faa71bb198ee34ae4893e4610aa83773645

Change-Id: Id87af31e9b60d548959ba735b635ff3d154fc587

Change-Id: I1ece738e7f60b6bbbdc802339f8b9deec9396e1b
Reviewed-on: https://gerrit.libreoffice.org/12469Reviewed-by: Caolán McNamara <caolanm@redhat.com>
Tested-by: Caolán McNamara <caolanm@redhat.com>

It takes a lot of time and we can save ~1% of size if at all. Rather
store them uncompressed in the zip file.
With png, we can save a bit more, although it's still only about 4% - 8%?

Change-Id: I43a3e3812882e4ce85e7af9d0aaab454d98c4860

Change-Id: Iae804a34f344aa793a6d5c13315f7bc1eb64c0a2

Change-Id: I66cbbfb2aa6abc6c8ebe34d9ea69855436c23edd

Change-Id: Ic2538ca8b0f7261064e1dfbf3884dd452003c797

Change-Id: I4d98b6f8b3315b731206700eb65f08463299dda3

For that:
1, create ZipPackageStream::successfullyWritten to be called after
the content is written
2, Do not take mutex when reading from WrapStreamForShare - threads should
be using different streams anyway, but there is only one common mutex. :-/

Change-Id: I90303e49206b19454dd4141e24cc8be29c433045

Preparation step to parallel deflating.
Rename putNextEntry to setEntry and make it a static function.
We need to call setEntry before starting thread but writeLOC after.

Change-Id: I99a9ffa7dc4c18b47c621847b48bf8469bfb789a

Change-Id: I504f5c0c9aa9b655ffb53d9820a33677dad6aa08

Change-Id: Ib8fa3351ba25416a13d6c8bf63bd5fc8e43703c5

Change-Id: Ibf81dc3cd8a9a9da3dfd6ee6e587a522c4d56a44

Change-Id: Ib6a69a83f4a378df838b2231b9eba7fba49cd9f1

Change-Id: Ie3f8ac261a70c9a2b5182fc7d36938d0a46ec045

We want to use ZipOutputEntry only for deflating (and maybe rename it).
ca13a937 was not a good idea because the
data still needs to be written sequentially anyway. Otherwise it's hard
to get offset positions of individual entries right.

Since this commit rawCloseEntry needs to be called always; also when we
use write&closeEntry because we don't call writeEXT in closeEntry
anymore.
Need to rename and add comments later.

Change-Id: I03bd48ca6e108e6253a77a137746165909ca3c3d

Since commit 0c5bb3f4.

Change-Id: I5511f9dc829aca42790f05cb8fb3ebd83b2acad4

one can find callgrind output file more easilly

Change-Id: Ic5019d5259a63ab23ed44d39baaeb7432690382e
Reviewed-on: https://gerrit.libreoffice.org/12450Reviewed-by: Riccardo Magliocchetti <riccardo.magliocchetti@gmail.com>
Tested-by: Riccardo Magliocchetti <riccardo.magliocchetti@gmail.com>