Change-Id: Ie2cdbf1291ec340053238b72dd155f20aa2aaddb

Change-Id: I665bf0bdb4e0d4c104a08f37a8afa9de4ad4a9f4

Change-Id: I82735e3f09f22763ef48aab18dee867159dbf842

Change-Id: I0d0d6e7ef42b3fbddbb8f4267099ea3b029c01ac

The RTF spec says control words are case sensitive, and page 252 of the
v1.9.1 spec defines \mlimloc, but earlier page 118 and 125 also defines
\mlimLoc, so let's accept both.

Change-Id: I01c9a166d0faabed4a92a0aa42a123432cb7e847

Change-Id: Id432bec2ad19b0ff71135cbe65f6e8ba42bf036d

Change-Id: I25394e009009ae9c111e929d493eb949c042e1bc

Change-Id: I50d55f5e5cc7987550ade689ca3754c818efeb62

There seems to be no obvious reason why we would need to store the chart
doc during xls import. The chart test docs don't show any difference
without this call.

This speeds up xls chart import quite a bit. With the test doc from the
bug report I'm able to open the doc nearly instantly now.

Change-Id: I0b80462b9d1c405ab1d4cf456e12ebfe4428db98

Change-Id: I94bf89867cec9cc24f8b96635c515d9477b561a9

The font cache in VCL returns different fonts for identical parameters,
which causes layout differences in Writer.

first we search the font with language 1054:
 debug: XXX 0x2103950   add to maFontNameList: TH SarabunPSK->waree
 p maFontNameList
 $9 = boost::unordered_map with 1 elements = {
   ["TH SarabunPSK"] = "waree"
 debug: XXX 0x2103950 found in maFontNameList: TH SarabunPSK->waree
in this very same invocation the "waree" is then rerouted to Deja Vu
(which can only be because it was called with language 1033):
 debug: XXX 0x2103950   add to maFontNameList: TH SarabunPSK->dejavusans
 p maFontNameList
 $11 = boost::unordered_map with 1 elements = {
   ["TH SarabunPSK"] = "dejavusans"
 }
 debug: XXX 0x2103950 found in maFontNameList: TH SarabunPSK->dejavusans

- the selection of different fonts happens in
  FcPreMatchSubstititution::FindFontSubstitute.
- we never get a direct cache hit from maFontInstanceList because there
  are some differences like WIDTH_DONTKNOW, and the maSearchName
- the first time a font is searched, it ends up in maFontInstanceList
- the second time a font is searched, it may end up in maFontNameList,
  which maps _only_ based on the name (appears to be an invalid optimization)
- once we have TH SarabunPSK->dejavusans in maFontNameList we can never
  select the other one because apparently dejavusans covers all languages

Change-Id: Ibd73de88d8fc3b6e1319eb34c261e55ea217a988

... to actual minimum so that the fix for fdo#33090 in
UniscribeLayout::GetNextGlyphs subtracts exactly the right number.

The original value of mnSubStringMin is guessed and may be up to 8
smaller than the actual start of the first VisualItem.

For the bugdoc it happens that sometimes it's too small by 8 and then
the wrong entries in mpGlyphs2Chars are initialized and then apparently
there are single glyphs that represent multiple characters and so
GetNextGlyphs returns a -1 character which is invalid.

 ->GetNextGlyphs dir 1 36 mnSubStringMin 28
 ->GetNextGlyphs g2c #1 [8] = 36
 ->GetNextGlyphs g2c #1 [9] = 37
 ->GetNextGlyphs g2c #1 [10] = 38
 ->GetNextGlyphs g2c #1 [11] = 39
 ->GetNextGlyphs g2c #1 [12] = 40
 ->GetNextGlyphs g2c #2 [4] = 40
 ->GetNextGlyphs g2c #2 [3] = 39
 ->GetNextGlyphs g2c #2 [1] = 38
 ->GetNextGlyphs g2c #2 [1] = 37
 ->GetNextGlyphs g2c #2 [0] = 36
...
 ->GetNextGlyphs init nCharPos -1
 ->GetNextGlyphs g2c [2]  nCharPos -1
 ->GetNextGlyphs set pCharPosAry -1
layout[0]->GetNextGlyphs 768,1024 a1800 c-1 0

Change-Id: Ie33ec797a412aa898bec3f4e8f97b88dcfed4d11

the general goal is

1/ to avoid the bloat of pet conviniences rules that
can be managed using other git provided means

2/ to work toward the goal of a read-only-source-tree build

Change-Id: I5ff9836e135507dcc53b37fd1098aa7876612bfc

Change-Id: I49d81f482c6b97627e6e7cc3741246b70383eb5a

Change-Id: I410ab7bbd74b616fcb0a3ed3b0c917cebeac23c6

Change-Id: Idf628ad51ebe9c841090a5e04c7986c15c726480
Reviewed-on: https://gerrit.libreoffice.org/1463Reviewed-by: Olivier Hallot <olivier.hallot@alta.org.br>
Tested-by: Olivier Hallot <olivier.hallot@alta.org.br>

Like the minimal filter tests, these tests are invoked from a single
cppunit method since commit 0a30cb22, so
some info about exactly what file failed is necessary. If we are at it,
also print the (re)loading time as well.

Project: dictionaries  f0c914a43e7e6540300da25c935a77aebb672094
null
Brazilian spell checker and grammar checker update.

Change-Id: I8967a353bbac45f9ea29cbc449e60fbf63f287a0
VERO: Version 3.0.0 beta 24
Reviewed-on: https://gerrit.libreoffice.org/1459Reviewed-by: Olivier Hallot <olivier.hallot@alta.org.br>
Tested-by: Olivier Hallot <olivier.hallot@alta.org.br>

Change-Id: I7313cca3335e1b41df2a70fe968bc1e6386a1e36

Some screwup with hypot(). As a workaround, include <math.h> early.

Change-Id: Ic8c6abc45f8502097c6b6b619108d50ea3c22392

There does not seem to be a point to using CFLAGS here, usual convention
appears to be to use AFLAGS.  Not target local, since right now there is
no need for that.

Change-Id: I5d2f45777c226c0afc632adc21e446f09740d75d

capitalType functions from the hyphenator, thesaurus and spellchecker
dispatcher have been replaced by a unique function in linguistic/misc.cxx.
The capital type is now represented as an enum.

Change-Id: I0c2768b06f995cf34c105de117c7bab138c727a9
Reviewed-on: https://gerrit.libreoffice.org/1460Reviewed-by: Eike Rathke <erack@redhat.com>
Tested-by: Eike Rathke <erack@redhat.com>

Change-Id: Iad4deb829f71a9824e881d9f5782a474167d7a28

...this was a transitional hack to get XML-format service.rdbs in.  Now that
registry-based bootstrap_InitialComponentContext is gone, XML-format .rdbs need
only be handled in cppuhelper/source/defaultbootstrap.cxx (so the
textualservices stuff once duplicated to there now effectively moved there).

Change-Id: Ifb93558768095c1b462fe4057ebf8724968cca77

... com.sun.star.comp.helper.RegistryServiceFactory (juh.jar).  Superseded by
default bootstrap mechanisms.

An aborting stub for non-inline cppu::createRegistryServiceFactory is left in
cppuhelper/srouce/compat.cxx to avoid having to incompatibly change
cppuhelper/soruce/gcc3.map.

Change-Id: I590e50b8f57e86d4bb3e00d157c9e5907c02f267

Change-Id: I27fd4d1f1dfafccaa7936defb5cd687d6266e0a2

...instead of deprecated cppu::createRegistryServiceFactory.

Change-Id: I64431f3d404e24267953de3fb94e3e5241db30a8

Change-Id: Ie65482f328a0932c3543200d1906940b1eb46897

Change-Id: Icb8ccabba162a838187d737ff30888f64dd695b6

Change-Id: Ic62056e2df9a17fd6716ae90e4372dec1d5dd40f

Change-Id: I433512dbd533fe6c84e94059f378c4a7a6e8671d

Change-Id: I6fc53e185254a37df7d03ab6f91210b8e397db2e

Regression from d4069372

Should be good enough for now, at least doesn't break any existing
functionality if you don't use the system provided full-screen
buttons. Should continue to compile against older SDKs but still work
also on newer OSes.

There are some known glitches. For instance if you choose View:Full
Screen, the window does go into the system full-screen state, but if
you then use the system unfullscreen button on the (now no longer
disabled, but hidden) menubar, LO keeps displaying its own small
floating unfullscreen dialog...

I also occasionally managed to get some menus in the menubar borked
with no text!? But as traditionally a full-screened LO has no menubar
at all, technically that isn't a regression;)

Partially based on my own old patch, partially on
http://svn.apache.org/viewvc?view=revision&revision=1423520 ,
http://svn.apache.org/viewvc?view=revision&revision=1423523 by Herbert
Duerr <hdu@apache.org>.

Change-Id: I546c58f7cee9f8cb6c746eb014d8d5a8af3619cb

reset members before calling GetBitmap(), as apparently they are used by
GetBitmap(), resulting in missing/not rendering parts of the preview
image in the bugdoc.

(regression from c0ce7ca4)

Change-Id: I02a6abb822900e1a28a1c632a122c1e093b73553

Doing word-count by switching per-word between two different
kinds of word iterator was insanely slow. This preserves an
ICU break-iterator for each type of word-breaking.