Generalized the OSX ppc workaround for ancient bison versions,
factored out for all gcc platforms. Put bison version detection
into configure accordingly, to switch on that, and not on platform.
Otherwise we would fail to import the cell contents of those documents
that don't include table styles at all. Some hand-crafted ods documents
don't provide table styles, which 3.4 imported just fine.
Older releases stored a few function names not defined by ODFF, namely
EASTERSUNDAY instead of ORG.OPENOFFICE.EASTERSUNDAY, TDIST instead of
LEGACY.TDIST and B instead of BINOM.DIST.RANGE.
Since OOo/LibO 3.3 the proper function names can be read, additionally to the
"wrong" names. Now it's time to write the proper names and still accept the
incorrect ones.
Test cases are attached to AOOo issues:
ORG.OPENOFFICE.EASTERSUNDAY
https://issues.apache.org/ooo/show_bug.cgi?id=112882
LEGACY.TDIST and BINOM.DIST.RANGE
https://issues.apache.org/ooo/show_bug.cgi?id=110229
Note that the FALSE in A2 and A3 is a result of the string comparison of the
actual formula, that differs in separators (, vs ;) only.
(cherry picked from commit a9b03bd199)
This is also another unfortunate bug due to the logic change caused
by the DECLARE_LIST removal. There was one crasher due to out-of-bound
array access, plus one incorrect behavior concerning cell content
placement. Both are now fixed.
* on OSX 10.5 PPC, ld crashes with a bus error, presumably because
of the empty ar[] string array.
* the "size = 0" default case looks wrong, replaced by *size=0
Some places in the code assumed that if the stack is empty a null pointer is
returned by top() (or old Pop()), this doesn't work anymore with ::std::stack
that instead has undefined behavior in that case, so check !stack.empty()
first before accessing top.
(cherry picked from commit ac40f7d650)
gsoc input bar now (should) snuggly wrap the text when collapsed, if multilines exist then scrolling with the keys should be clean and no spill from upper or lower lines visible. Removed some strange ( pseudo padding ) apparently for taking into account the window border and reorganized the resize logic so the padding is added ( and commented why ) in just one place.
