Nikita Kniazev

``` >py -3 -m clcache C:\Users\username\AppData\Local\Programs\Python\Python35\lib\runpy.py:125: RuntimeWarning: 'clcache.__main__' found in sys.modules after import of package 'clcache', but prior to execution of 'clcache.__main__'; this may result in unpredictable behaviour warn(RuntimeWarning(msg)) ```

Autodiscovered Python will be used for building even if it is of different bitness

2

32bit Python is installed in `C:\Python27`, but `b2 address-model=64 toolset=msvc-14.1 python` tries build it, and of course fails: ``` C:\Python27\libs\python27.lib : warning LNK4272: library machine type 'x86' conflicts with target...

warning: type ‘struct _stack’ violates the C++ One Definition Rule [-Wodr]

1

Building B2 with `-flto` uncovers a bug: ``` > g++ -x c++ -std=c++11 -O2 -s -flto -DNDEBUG builtins.cpp class.cpp command.cpp compile.cpp constants.cpp cwd.cpp debug.cpp debugger.cpp execcmd.cpp filesys.cpp frames.cpp function.cpp glob.cpp...

variant=release runtime-debugging=on does not enable assertions as stated in documentation

1

The `runtime-debugging` says: > Specifies whether produced object files, executables, and libraries should include behavior useful only for debugging, **such as asserts**. Typically, the value of this feature is implicitly...

Travis is configured to build Boost.DateTime every week, the build takes 12 hours. Please consider configuring ccache. *Update: Previously overlooked and thought it is configured for every day job*

X3: Fix inconsistent error position for `p` vs `p >> eps` or `repeat(1)[p]` in error_handler

5

* Do not rely on iterator rollback on fail: Preparation to relax iterator rollback policy, which is also currently violated by not rolling back skipper almost everywhere, so this should...

X3: Refine roll back iterator policy

13

A long time ago I had an idea about rich and low overhead error reporting and it did not work out because parsers roll back iterator in some unexpected places...

```cpp #include int main() { using boost::spirit::x3::int_; char const* const it = ""; int i; parse(it, it, -int_, i); } ``` The problem is that after a successful parsing the...

In #607 I rewrote the alternative parser attribute deducer to not use MPL, and while it is equivalent to the original, I forgot that the original was very questionable. It:...

https://github.com/boostorg/regex/pull/116 removed `boost/regex/pending/unicode_iterator.hpp`, we need to find a replacement or bundle it.