Doxygen would probably recognize these without () too but add them for
consistency.

sed -i "s,<member>\([^<]*::[a-z][^<:]\+[^)]\)</member>,\1(),g"

Change-Id: I2615b99265b75633459e35164e54d9da7fe76b85

Doxygen will only recognize a un-qualified method name as such if it is
followed by "()".

sed -i "s,<member>\([a-z][^<]\+[^)]\)</member>,\1(),g"

Change-Id: I69bc17849e76f3a3d91c6daf0f1be8168a83cfc5

This looks much better now.

sed -i 's,<type scope="\([^"]*\)">\([^<]\+\)</type>,\1::\2,g'

Change-Id: I94cd0f93afa89855b62dadeb229d2b2e1775cd80

Doxygen does not know type element and will recognize strings that
contain capital letter (all API types do) automatically as type.
This patch removes 15k doxygen warnings.

git ls-files | grep \\.idl | xargs sed -i "s,<type>\([^<]\+\)</type>,\1,"

Change-Id: I45c07cf0b115d5fb5353f4aa9719839615ea1150

sed -i "s,<type>\([^<]\+\)</type>,\1,"

Change-Id: Ibdaeb2c396ad29ced05716142b38c6a1c3e49847

sed -i 's,<type scope="\([^"]*\)">\([^<]\+\)</type>,\1::\2,g'

Change-Id: I7eb32acd44132de79cc7ac1e7f78b43bb4814d14

Unsurprisingly doxygen does not support reading a idl_chapter_refs.txt
file like autodoc does.  And it looks like doxygen tag files can not
be used to do this, as they link only in the "other" direction.

But with a bit of hackery we can generate a "dummy" IDL input file, with
emtpy definitions for those types that have DevGuide links.
Doxygen will nicely combine documentation from multiple definitions of a
type, and fortunately does only put in references to the first file that
defines a type, so hopefully adding the generated file as the last input
file is suffficient to make things work nicely.

The generated file is committed to git because there is a bit of a
problem with finding out what sort of entity a name like
"com.sun.star.Foo" refers to, which needs to be in the doxygen input.

Change-Id: I793bc59c8acfecf460e4addea7838c76a3ede77e

Change-Id: I04b8dbecbe62f8d72ca75f6508046294c3409a17

Change-Id: Ibf7ab46bdd08a42ecf475500d35f92b67d4d8036

Change-Id: I9a90bf06a70f6537f9671ef2cd47bf68c3cf5d57

Change-Id: Ic3f2d42121b5423d67c4c79302b02ac495ac9640

- give right path to STRIP_FROM_PATH
- prefix vars
- depend on makefile
- only call cygpath in rule

Change-Id: Iceef191751b82a423be354db1f0a3ca4fc213e04

Change-Id: I273b019409f36d0b4488458e152099493ce81147

Change-Id: I9c301f7d8b5295854e12585fa12df416ed5dbf02

Change-Id: I121f0d3d646ce434ea1451a94a5c911005dc21f9

Project: help  efa1b1723eae3e378fbd299677745be1364e53ce

Change-Id: I0d62ae391b12cec9a057aef26344f3454d2faba4

Change-Id: I9662671d8dbce87993ba67cf8e4fd993ad1eb27a

Change-Id: I746e1fa25137db786f82dda50bfa4b497e94cefd

Change-Id: I2ad35bfbaf40dd1a5b5c9f7f119268b8f30ca093

Also, use a better way to fetch a double array.

Change-Id: I65a5bb9c972620f63e79a84b461245cc7fea1a51

Change-Id: I790db61d2a60cf4074fd2e3291ec42f0a4649b9e

Change-Id: Id9009faa5c9221a04130f87af4fa14e33fe0dc50

Change-Id: Id735c7ef89411581f7bc8d7f10b3edfe8aba0c72

When a fly is anchored at the first paragraph of a ToX, as happens in
the WW8 bugdoc in i#30979, updating the ToX will remove the anchor
node, and bad things happen, including ~SwIndexReg assert and also
eventually a crash when editing the Frame and updating the ToX a few
times.  Probably best to really remove all anchored flys here.

Note that it's not enough to re-anchor to the empty node that is
inserted here because that one will also be removed again.

While we're at it the _DelBookmarks is superfluous because CorrAbs
does exactly that 2 lines above.

Change-Id: Id40ae1d88ad99813b93bea7470f5dc790edcc80c

It have been unused on Mac OS X since since 2008 (commit
da4d070d).

Change-Id: I4b3eabe71a54021903009310a87319c9caecfd4c

Change-Id: I660a895a227da3d28a0b410c56f2395961a5bc4f

This file is Core Text only already.

Change-Id: Ic4c1f44296eaae582091e515d3ca1b4aa227301e

When using a graphic filter like blur or sharpen, a raw uncompressed
bitmap is created. Until now it was not possible to rotate such bitmap
but now this is allowed.

Change-Id: I716fa46e37680b1111131586fad20da28391de14

DialControl internally stores the angle to a 100th of a degree
percision, but when SetRotation is called, the value was always
rounded to a 1 degree percision. This has now been modified to
round only on DialControl mouse interaction.

Additionally add add a parameter to SetLinkedField which states
the number of decimal places the NumericField is set to. This change
makes possible to set the NumericField to 100th of a Degree as the
internal angle is represented in. Default value is 0, so no change is
needed to existing DialControls.

Change-Id: I1f0c20092cdfccbd0878d7eb620bfdad7825b0fb

Change-Id: Ie57cf12a3dc0984367c34dc4f52fd00036cf3d9d

Change-Id: I60f92dd13282281d77571e41707d68ae55722d9b

There is some strange handling for row 0 in the calc core code. Without
a very deep inspection I can't decide whether the test or my change is
the problem but for now we can live with the old behavior.

Change-Id: I69615349f5b46856aba36dada235a49c84eb03af

Change-Id: Ide65a35d1dd39cbce9230f951bcc229b0df7b634

Change-Id: Ied47c5268b4554b219028281916d884aebf816dd

Change-Id: I1310ac912aa2031b15851339fd84d435308350a9

Change-Id: Iac028806fd306253fe0e4b407302ff66f853b49c

Change-Id: I01c6681da8a0ac83eaafdf86a04a187f13a05792

We only support grouping of formula cells in the column direction, and
skipping is not supported. Let's reflect the test cases for that.

Change-Id: I0229b90b7d9a2eea29d6eb2ec2b91e63d760a352

Regression from cbd1a896, that commit
accidentally not only removed the #if 0 block, but also two other
braces -- because of this, in case bStorage was false, nothing was
saved. The effect of this was that if the user edited some macro in the
basic IDE, then saved the macro, it was shown as saved, but in fact that
didn't happen.

Change-Id: I20f8358dddfb226976be72f95dcad64de88d83c3