First only parts of the parser state was copied, then later the whole
state, but code not using the full copy was still there. Remove it now.
Change-Id: I2c5507e74a24b8dac74552c61d69ba0be7257d6b
This is especially the case when the source is a (saved) query.
Fixes: Report Builder wizard "Finish" button does nothing -> cannot create report through wizard
Change-Id: I266772035435a7c623beb0c0f66fc88e6316be8e
+ hyperlink starting before a field and ending in its results isn't
valid.
+ when footnotes have hyperlinks we need to add the proper XML
namespace and output the relations to a different file.
Change-Id: I1c6cb8acba21c35e10bb0497eeaf96bbd73730d1
This pItemDesruptList is pointless now. We're registering the delete-when-idle
items in their ctor's and deregistering in their dtor's which get called from
the idle callback. The idea of pItemDesruptList appears to be for use in
deleting any items at exit time whose idle callback didn't get called already.
Which was done by DeleteOnIdleItems which itself was removed a while ago
because nothing was calling it (for years). Some experiments in making it a
simple singleton and/or one that hooks off default component dispose or
XDesktop dispose shows that its crash city, so lets just remove the unused
stuff outright rather than trying to fight to get some 12+ year unused appendix
used.
Change-Id: Ie0256d6987cf89a2a12db297065af09674547b3e
This reverts commit b8f7cc3c26.
This does not work, because there is no dependency between the
*Object .d files and the *Objects, and so the concat-deps races with
the compiler writing the dep files.
Introducing such a dependency would of course be reverting
8b5a984d45
A blacklist based approach doesn't seem easy since there are various
different SDKs, MFC, ATL and whatnot each potentially with an own
include path; so simply whitelist the known source and build dirs.
Change-Id: I93b27bd0ffbd57237da2d23afa95dea7c0785774
* in some scenarios gcc (or ccache) generate relative paths
* thus we need to concat the depfiles in the same makerun as the compiles
* stiil not foolproof:
- cd sw && make and abort with Ctrl-C
- cd .. && make sw
- now we have deps from mixed workdirs
* in the end we should likely always cd to SRCDIR before all compiles
Change-Id: I007da96f38ed2cfe403b692400f447764f6d6988
remove extending of pixmapRect in the nPart == PART_WINDOW branch
and simplify code, since rControlRectangle, pixmapRect and widgetRect
are the same.
Change-Id: I253851fad558e525cd533bb3576a9bba82187e53
This reverts commit 3e3acee762.
It was a nice idea, but C++ is not yet ready for it; with the travesty
of parametric polymorphism in C++ the find_unique inside the definition
of find_unique actually refers to find_unique<Value, Compare>, so there
is no way to actually refer to template<Value, Compare> find_unique
inside its definition. Thanks to Luboš Luňák for explaining
the problem to me. Somehow this does work in GCC 4.7 even with
-std=c++98, likely by accident.
To get footnotes sorted properly one need to:
* actually use the comparator class
* compare on the offsets value (and not there pointers)
Change-Id: Ie86802c7a92602bfbb4d2a603afe06f621e392a3
8cd05e9cf1 changed from using String to
OUString. ScCsvGrid::ImplSetTextLineFix() attempted to copy excess
characters (always CSV_MAXSTRLEN if greater than field width) where
String::Copy() silently ignored the excess length but OUString::copy()
may result in invalid memory accesses and asserts in dbgutil build.
Change-Id: Ic9f7f38d6f2bbd770d6356e1304de8e39c09e30b
This reverts commit 51e3df40bc.
The fix was correct at the time it was done, but unfortunately now it
conflicts with 89f08dce89, which fixes the
same problem by changing the definition of DFF_ANIM_PROPERTY_ID_COUNT.
