My previous commit was wrong:
| commit 28aed34f517109816ba7950aa134ee3260d82dbd
| Author: Laurent Rineau <laurent.rineau@cgal.org>
| Date: Fri Aug 8 15:59:29 2014 +0200
|
| When GMP is disable, use LEDA types instead
|
I stupidly forgot to quote ${CGAL_DISABLE_GMP}. Without the quotation,
regardless of whether the variable was empty or not empty, the test
[ -n $var ] returned true.
Distribute dijktra header file from boost 1.55 to work around
a bug in the version 1.54. The header file is only used if the
user compiles with boost 1.54
Tested in CGAL-4.5-Ic-102
Approved by the release manager
Conflicts:
Installation/changes.html
Reintroduce the possibility to use CGAL without GMP
(to use LEDA number types instead for example).
The cmake option to set to switch off GMP when building
the lib is CGAL_DISABLE_GMP.
Tested in CGAL-4.5-Ic-86
Approved by the release manager
ge is necessary,
The error was:
/usr/local/include/boost/type_traits/has_new_operator.hpp: In function 'bool test_ray(Arrangement_2&, Face_handle)':
/usr/local/include/boost/type_traits/has_new_operator.hpp:24: error: 'template<class U, U x> struct boost::detail::test' is not a function,
CGAL-4.5-Ic-86/cmake/platforms/x86-64_Darwin-13.0_Apple-llvm-gcc-4.2_Release/test/Arrangement_on_surface_2/test_unbounded_removal.cpp:24: error: conflict with 'bool test(const Arrangement_2&)'
CGAL-4.5-Ic-86/cmake/platforms/x86-64_Darwin-13.0_Apple-llvm-gcc-4.2_Release/test/Arrangement_on_surface_2/test_unbounded_removal.cpp:55: error: in call to 'test'
I do not understand how `boost::detail::test` was introduced in the
matching set of functions. The `boost::detail` namespace is not supposed
to be involved.
bad() implies fail() but std::basic_ios::operator! is actually equivalent
to std::basic_ios::fail(), not bad().
For example,
std::ifstream is(argv[1]);
if (is.bad()) {
is probably wrong because, in case the file does not exist, only the failbit it set,
and not the badbit.
Other example:
double x;
is >> x;
If the stream 'is' does not contains a string that can represent a double,
then failbit is set (and maybe the eofbit if the stream is actually too
short), but not the badbit.
bad() can be used, in case fail() returned true, to discriminate between a
I/O error and a logical error.
See:
http://en.cppreference.com/w/cpp/io/basic_ios/operator_boolhttp://en.cppreference.com/w/cpp/io/ios_base/iostate
* Enhancing the testsuite---adding tests for
batched, point location
vertical decomposition, and
triangulation point location.
* Various bug fixes:
bugs in point location.
memory leaks found by Andreas.
* Resolving several small issues to remove compiler warnings.
Tested last in CGAL-4.4-Ic-54.
Approved by the release manager
Andreas Fabri
Efi Fogel
Laurent Rineau
Sébastien Loriot