[PUSHED (mostly) - RFC - Solaris/intel/gcc patches

[Michael Meeks]

Hi Pierre-Eric,

On Tue, 2012-07-10 at 15:46 +0200, Pierre-Eric Pelloux-Prayer wrote:
> Here are several patches needed to build and run LO on Solaris/intel/gcc.

	Lovely :-) thank you !

	I committed all except 0003 and 0006 which need some gnumake expert I
think, as you say 0008 is not really what we want - instead we want the
unit tests to run everywhere unless there is a good reason why not :-)
if they can run on Windows, Solaris should be easy - right ? ;-)

	So - overall, it's great to bring the Solaris port back from the dead,
and without all those horror #ifdef for compiler oddities using gcc.	

	Bits like the -pipe removal in com_GCC_defs in 0003 looked like we
ought to make that a platform conditional, turned off on solaris.

+$(eval $(call gb_Library_add_ldflags,PresenterScreen,\
+    -L$(gb_Library_OUTDIRLOCATION) \

	looked odd in 0006 too - but overall this is a great set of fixes :-)

>    * some .so files don't have the .3 suffix, but the generated install 
> script : ./solver/unxsogi.pro/bin/setup_osl.ins is looking for .so.3 
> files. So manually editing this file is needed.

	That is odd. On linux at least the ure/lib libraries that we would link
to are .so versioned with a .3 in every case - I wonder why it fails to
work for solaris.

>    * to start LO, you'll also need to specify a LD_LIBRARY_PATH like this:
> LD_LIBRARY_PATH=./solver/unxsogi.pro/installation/opt/ure/lib/ 
> ./install/program/soffice

	and that's odd too; soffice is itself a shell-script (much as I'd love
to kill it), possibly you just want to add Solaris to:

case "`uname -s`" in
NetBSD|OpenBSD|FreeBSD|DragonFly)
# this is a temporary hack until we can live with the default search paths

	one of those rather long-term temporary hacks ;-)

> These are not enough to get LO running properly because it still
> crashes after some time (see http://pastebin.com/JKQHnt3Z)

"terminate called after throwing an instance of 'com::sun::star::uno::RuntimeException'"

	This is a -particularly- annoying "something went wrong" message - and
-worse- it is reasonably trivial to have it print out more useful
debugging information: we usually know much more precisely what is
wrong.

	If you hack desktop/source/app/sofficemain.cxx - and un-protect the
try/catch guarded by ANDROID there (except for the random /assets ini
setup) then life will prolly be better for you wrt. a nice warning
message printed.

	Thanks again !

		Michael.

-- 
michael.meeks at suse.com  <><, Pseudo Engineer, itinerant idiot