2011/1/29 Rene Rivera <grafikrobot@gmail.com>:
On 1/29/2011 12:38 PM, Joachim Faulhaber wrote:
Dear release managers,
as discussed before, I've done my homework related the bug that I've detected, which is described in ticket #5135: https://svn.boost.org/trac/boost/ticket/5135. See details below.
2011/1/27 Joachim Faulhaber<afojgo@googlemail.com>:
2011/1/27 Rene Rivera<grafikrobot@gmail.com>:
On 1/27/2011 10:48 AM, Joachim Faulhaber wrote:
Such last minutes surprises are always not very welcome, I know ... I won't take your decision personal ;)
After a brief talk on IRC.. How about if you..
1. Add a test to show the problem in trunk. Passing with the fix in it of course.
1.1: I've fixed the code (see patchfile) Index: is_interval.hpp =================================================================== --- is_interval.hpp (revision 68546) +++ is_interval.hpp (working copy) @@ -26,7 +26,7 @@ { typedef is_interval<Type> type; BOOST_STATIC_CONSTANT(bool, - value = (interval_bound_type<Type>::value< interval_bounds::undefined)); + value = ((interval_bound_type<Type>::value)< interval_bounds::undefined)); };
@@ -34,7 +34,7 @@ { typedef has_static_bounds<Type> type; BOOST_STATIC_CONSTANT(bool, - value = (interval_bound_type<Type>::value< interval_bounds::dynamic)); + value = ((interval_bound_type<Type>::value)< interval_bounds::dynamic)); };
template<class Type> struct has_dynamic_bounds ===================================================================
1.2 I've added two test to the trunk that test the bug fix:
(1.2.1) A minimal test containing the bug and it's fix
http://svn.boost.org/svn/boost/trunk/libs/icl/test/fix_include_after_thread_...
(1.2.2) A test with includes of ICL-files that caused the error before when preceeded by boost/thread.hpp
http://svn.boost.org/svn/boost/trunk/libs/icl/test/fix_icl_after_thread_/fix...
The tests ran successfully for 2 nights.
2. Wait for the beta to get done.
Has happened tonight.
3. Then you can ask again to make this change after the beta is out. And we'll likely say yes.
I have prepared a merge that does the fix and adds the two new tests. Do I have permission now to commit the merge to the release branch?
I'm going to say OK.
Done!
One worry I have
Don't worry . . . beee happy :)
is that some of the release toolsets haven't run that test yet on the trunk :-( But there's nothing you can do about that. And they are the older set of compilers anyway (i.e. msvc8).
The fix is extremely trivial.
And thanks for being patient with us :-)
It's been a pleasure for me :D
PS. It seems some toolset should be marked as N/A as they are failing most of the ICL lib tests.
You're right. I should do that for sun and borland. Concerning clang I think it's a compiler error, that the clang people should be interested in, so N/A would hide that information.
PPS. Yes, I'm breaking my own no-work-in-the-weekends rule :-(
At least I appreciate that :) Thank you again, Joachim -- Interval Container Library [Boost.Icl] http://www.joachim-faulhaber.de