Age | Commit message (Collapse) | Author |
|
Change-Id: Ifebd5ad91d1c8b456c458b01738261dfe34a6f83
|
|
Change-Id: Ibc4034cf4eb2d610fdb6dcabb3d547bf2f9f1172
|
|
|
|
callgrind (on text/swriter/guide translated in 'fr')
before Ir = 889,687,925
after Ir = 406,372,177 (405,668,292 w/o the 'native' memory allocator)
time confirms:
before 0.288/0.272/0.016
after 0.146/0.131/0.015
the output before and after were diff-ed and every generated xhp files
are identical.
All improvements where done still using 'normal' sal and c++ API.
There are still some low-hanging fruits, like XMLUtil:QuoteHTML
(probably 20-30 millions Ir to shave there)
The destruction of XMLElements is also singularly high
(11% of the remaining Ir count)
But the bulk of what is left is OString management (alloc/acquire/release/
free/copy), and I/O using streams
Change-Id: Ia316c953cd4bd46fc33a58a0c10f26b0ffa042c2
Reviewed-on: https://gerrit.libreoffice.org/1262
Tested-by: Norbert Thiebaud <nthiebaud@gmail.com>
Reviewed-by: Andras Timar <atimar@suse.com>
Tested-by: Andras Timar <atimar@suse.com>
|
|
Because either it's called from there, and then it's the current directory,
or otherwise the srcdir has no idea where the builddir is (or even how many
of them), and there's nothing to tell it.
Change-Id: I88c5fe8f015cc63a5aa7c327eb82ed751c3e8aa0
|
|
Change-Id: I2e93877283b0ea2a3b7162978a515d52396dfa72
|
|
Change-Id: Ie0d295cd7db695f777071edcd9b792056ac783ef
|
|
Change-Id: I7ac7392646ea34fb2e6a9f15990aca05a2c1e484
|
|
Change-Id: Ie4c934b52c4b010703e91f84b83e1fe590b5d187
|
|
Change-Id: Ia80a1246a7e3c8862a50a3daf79aba5609b7df3a
|
|
This patch allows to do
mkdir <build_dir>
cd <build_dir>
<src_root>/autogen.sh
make
Only the remaining dmake modules are poluting the <src_root>...
which will get eventually convererd to gmake...
Change-Id: Iefd64732fa12e096d554cff4eee6b777deb92338
Reviewed-on: https://gerrit.libreoffice.org/1273
Reviewed-by: Luboš Luňák <l.lunak@suse.cz>
Reviewed-by: Norbert Thiebaud <nthiebaud@gmail.com>
Tested-by: Norbert Thiebaud <nthiebaud@gmail.com>
|
|
Change-Id: Id82c9dc43d3af21d434ed5f9b7ef3bdae9c4289b
|
|
Change-Id: Ib1219e8e7031febd4f9cc599cb19426f974eac9c
|
|
Change-Id: Idf6003acc7ee07997150f77c886947dc1476218f
|
|
Change-Id: I4d33262fbff0c802379eea2c048048c85f4c373c
|
|
Change-Id: I30754e4d0cb0c490e6c125ba034e2e0eac3751e7
Reviewed-on: https://gerrit.libreoffice.org/1297
Reviewed-by: Michael Stahl <mstahl@redhat.com>
Tested-by: Michael Stahl <mstahl@redhat.com>
|
|
and hack up a mechanism to support menubuttons and map
the popups to the menubuttons
Change-Id: I6f3bd0e83b2c55f71aa85924fdc32d3a73ce0197
|
|
Change-Id: I0cbe7b4c6d329f2780cb3cac9bce6d722b7de469
|
|
Change-Id: I8d8dd139e57fb00decda8705157480439009cde8
|
|
Change-Id: I6767c62265dcfbb55ce086385802db59d4d0b177
|
|
Change-Id: I411133e81a9b548aaef2333c705c1376f45304d7
Signed-off-by: Marcos Paulo de Souza <marcos.souza.org@gmail.com>
Reviewed-on: https://gerrit.libreoffice.org/1294
Reviewed-by: Luboš Luňák <l.lunak@suse.cz>
Tested-by: Luboš Luňák <l.lunak@suse.cz>
|
|
Change-Id: I4912ffc42edbfa1456835e61163812c371eb14c6
Signed-off-by: Marcos Paulo de Souza <marcos.souza.org@gmail.com
Reviewed-on: https://gerrit.libreoffice.org/1293
Reviewed-by: Luboš Luňák <l.lunak@suse.cz>
Tested-by: Luboš Luňák <l.lunak@suse.cz>
|
|
- removed rtl:: prefix
- removed RTL_CONSTASCII_USTRINGPARAM
- corrected some misspells
Change-Id: I88bb0beec718a7fe38c61220aa61401419f23b42
Reviewed-on: https://gerrit.libreoffice.org/1291
Reviewed-by: Luboš Luňák <l.lunak@suse.cz>
Tested-by: Luboš Luňák <l.lunak@suse.cz>
|
|
Initializing the liblangtag database stuff takes much percentage of
startup time, avoid that.
Change-Id: Ieb12ff71c2b1421a5644acebaef4c8366694c249
|
|
For "simple" known (to us) locales avoid the overhead of liblangtag,
also in preparation of not having to initialize its database during
startup.
This also enables recognition of the 'qtz' pseudolocale for key ID
resources again, see fdo#57413
Change-Id: Id52a1ddc37b42063e3bf68a6dbeaeacfbfa704ef
|
|
|
|
and apple gcc-4.0.1 chokes on this
Change-Id: I4799d66f483f10d9b633077a16d25208c8b595e0
|
|
|
|
Change-Id: I861c49d2b0dfc57d6e777b14742cb324e4d4d0a8
|
|
Change-Id: I120acd40b83cad074869b404d4e74b5ec69d493d
|
|
Change-Id: I9d31602750dc6516b7bee5c0796328cee07b10ea
|
|
Change-Id: Iffa3296a15cbfaef8b7eca0648d9d954c92b3d4a
|
|
Change-Id: Ie8a2c7e473d717cc22c51205dd8cdc59934de8b1
|
|
|
|
|
|
|
|
Change-Id: I23b6b8fdbb9c4f367a590a21e40e7bbaa76f7fce
|
|
With MSVC2010 compile time goes 20 minutes -> 7 minutes. With Clang
it's however 7 minutes -> 10 minutes, apparently the overhead
caused by all the extra data is bigger than the gain.
|
|
|
|
Otherwise gb_CppunitTest_use_library_objects won't work with MSVC.
|
|
Rather than LinkTarget/external_headers, as this doesn't really make
any difference, except if PCH is made to also use headers from the same
target (not a very good idea for development, but might be useful
e.g. for buildbots).
|
|
Change-Id: I9041129289c24c4ee8a7a6d50e34b785969e8204
|
|
Change-Id: I6eb52f42c3364970553015492b9c488c04febed6
Signed-off-by: Marcos Paulo de Souza <marcos.souza.org@gmail.com>
|
|
This is to complement the RTL_FAST_STRING variant, which allows any
combination, so even two stringbuffers.
Change-Id: Ic15d81246b3d0f3e22b51c0516409e95c3a4dc8d
|
|
Change-Id: I692a9cfdca5ad50170eea2178a6aa87e8154b951
|
|
This rather substantially accelerates the first use of autocorrection.
Interestingly, it also appears to accelerate the sorting of the items;
potentially inserting sorted items into a set is a pathological balancing
case, that is avoided by the hash algorithm's randomness.
|
|
|
|
... objects at the right page. This was effectively disabled for all
anchor types except FLY_AT_PARA by a misplaced ! in the conditional,
and C++ not being a type safe language.
Change-Id: I64af2141e577ac896292eecd9f8c46c62a0548db
|
|
Change-Id: I097675450a0eed87eccf90356c81657485609d13
|
|
Specify subsystem version 5.01 in the linker options. Otherwise, when
built with the linker in Windows Kit 8.0, executables are created with
the "operating system version" and "subsystem version" fields set to
6.0 (which means Vista).
Attempting to run such executables on Windows XP fails with an error
that claims them to be "invalid Win32 executables" (which is a blatant
lie, of course).
We can well set the subsystem version to 5.01 as we don't
unconditionally use any APIs that wouldn't be present on XP.
Note that you need Visual Studio 2012 Update 1 to be able to create
code that runs on XP. The original version creates executables that
importx entries from kernel32.dll like GetTickCount64 that aren't
present on XP.
Apparently it doesn't hurt on XP that DLL headers specify a "too new"
subsystem version.
|