[GERRIT][PATCH] gbuild conversion l10ntools module

[David Ostrovsky]

Hi,

please review my gerrit patch request: 
https://gerrit.libreoffice.org/#/c/105/
The normal patch is attached (just in case you are more comfortable with 
it).

What has changed?

1.
The scanners are compiled as C++ code now:
cfglex.l
srclex.l
xrmlex.l
That why yyerror and yywarnings have changed their signature.

2.
flex options -w -l -8 are dropped.

3.
The generated scanners are not wrapped any more, so diagnostic pragma 
hacks are used on gcc to pass WaE configure option.
I tested it only on ubuntu (gcc version 4.6.1) though.

4.
HelpLinker & HelpIndexer are linked against dynamic librariy helplinker.
Well, I messed around with this one, so may be we have to clean up here.
For that i extracted HelpLinker_main.cxx and included new header 
HelpLinker.hxx in Package_inc.
I hope, i defined the programms on the right place in Repository.mk, to 
get the desirable rpath:
$(eval $(call gb_Helper_register_executables,OOO, \
     HelpLinker \
     HelpIndexer \
[...]

5.
-.IF "$(OS)" == "MACOSX" && "$(CPU)" == "P" && "$(COM)" == "GCC"
-# There appears to be a GCC 4.0.1 optimization error causing 
_file:good() to
-# report true right before the call to writeOut at HelpLinker.cxx:1.12 
l. 954
-# but out.good() to report false right at the start of writeOut at
-# HelpLinker.cxx:1.12 l. 537:
-NOOPTFILES=\
-        $(OBJ)$/HelpLinker.obj \
-        $(SLO)$/HelpLinker.obj
-.ENDIF

this was dropped.
Do we still need it? If uncertain, please activate it.

What has (unfortunately) not changed:
we still follow this ugly mixed case: HelperLinker and HelperIndexer
I don't know why.

Status pages is updated (status review):
https://wiki.documentfoundation.org/Development/Build_System/Module_status

Regards
David

PS
@Bjorn, thank you very much for helping with initial gerrit setup!