On 06/03/2013 08:42 PM, Ariel Constenla-Haile wrote:
[...]
Build is broken in three places (CentOS x86-64):
1) sal/inc/rtl/math.hxx:
../../inc/rtl/math.hxx: In function 'bool rtl::math::isFinite(double)':
../../inc/rtl/math.hxx:313: error: '__builtin_isfinite' was not declared
in this scope
../../inc/rtl/math.hxx: In function 'bool rtl::math::isInf(double)':
../../inc/rtl/math.hxx:325: error: '__builtin_isfinite' was not declared
in this scope
../../inc/rtl/math.hxx: In function 'bool rtl::math::isNan(double)':
../../inc/rtl/math.hxx:336: error: '__builtin_isfinite' was not declared
in this scope
dmake: Error code 1, while making '../../unxlngi6.pro/obj/math.obj'
In CentOS __GNUC__ seems defined, I see Jürgen added a fix for MACOSX,
so the logic of the defines might be simply not working at all.
The __builtin_isfinite() is used for for GCC as GCC's bug 14608 [1]
prevents the common isfinite() macro from working reliably. The builtin
should work according to the gcc documentation [2]: "GCC provides
fpclassify, isfinite, isinf_sign and isnormal built-ins used with
__builtin_ prefixed."
[1] http://gcc.gnu.org/bugzilla/show_bug.cgi?id=14608
[2] http://gcc.gnu.org/onlinedocs/gcc/Other-Builtins.html
Looking through gcc's sources the builtin was introduced in [3] from
2007-7-18 for gcc 4.3. I'm afraid we'll have to check for older gcc's
there and add another case for the ifdef GCC part.
[3] http://gcc.gnu.org/viewcvs/gcc?view=revision&revision=126725
Considering that isfinite() was in the C99, C++99, SUSv3, etc standards
for almost fifteen years now it is amazing how much trouble such a
two-bit function can still introduce.
2) graphite:
In file included from ../include/graphite/GrCommon.h:35,
from ../src/segment/Main.h:31,
from ../src/font/FileFont.cpp:16:
/build/aoo/src/clean/svn-trunk/main/solver/400/unxlngi6.pro/inc/stl/vector:31:33:
error: boost/tr1/tr1/vector: No such file or directory
Missing module dependency on boost.
Thanks! I added a dependency of stlport to boost.
Herbert
---------------------------------------------------------------------
To unsubscribe, e-mail: dev-unsubscr...@openoffice.apache.org
For additional commands, e-mail: dev-h...@openoffice.apache.org