GTest doesn't permit contextual conversion-to-bool until 1.8, so
explicitly convert to bool as a...

And mark the 'bool' conversion operator as 'explicit' when the compiler
supports it as that's th...

This may make it easier to refactor this code to possibly use namespaces
instead of a class.

Also remove -Weffc++ because the only warnings it produced were false
positives or not a problem.

Because apparently they don't recognize that option.

CHANGELOG update for cucumber/cucumber #251

Move json_spirit to a separate directory, out of the default
include-path. This so that we can m...

I.e. stop polluting the global namespace of our clients with symbols
from other namespaces just ...

It's basically just a namespace rather than a class. So we might as well
convert it to that.

Because StepCallChain's lifetime is always a subset of the StepInfo it
executes over. So just gi...

When we're asked for a step with an ID that does not exist, prevent
storing a mapping of that ID...

Because the base class forces us to create runtime instances of derived
classes, while the funct...

Also make it const-correct and use a vector for storing subexpressions.

We don't need shared_pt...

Url: https://coveralls.io
Implementation using lcov and gem coveralls-lcov
Fixes issue #123

For 1.6 GMock still has GTest as a dependency and so just specifying it
on the link interface is...

Because we can nicely incorporate version checking and version-specific
handling in that.

There's no polymorphism involved, nor shared ownership, so there's no
reason at all to return a ...

Because it already is. So make all its methods static to remove the need
for an instance to use ...

This was caused by QApplication's destructor being run after `main` as a
result of it being a gl...

Considering that the only difference is a single comparison macro: just
define that macro to be ...

This macro used to be necessary to prevent duplication of the function
argument list. This neces...

This closes #112

The hardcoded path to the Qt directory for OSX doesn't work (anymore)
because it refers to a non...

* Make building with Qt possible to disable (CUKE_DISABLE_QT)
* Enable building with Qt on our C...

That should fix appveyor build for mingw
+ small cleanups in examples CMake configuration

To ensure our Qt-using code gets tested by our CI systems as well.

* remove 'cmd' prefix
- because this is already the default
* remove intermediate RUBY_VERSION ...

Also use the more localized target_include_directories instead of
include_directories and AUTOMO...

Use the 'option' command for user-selectable options instead of cached
bool variables. Just like...

This gets rid of the FunctionSignature and FunctionArgs helper classes
by using a pointer-to-mem...

By producing JUnit output and submitting it to AppVeyor's job processing
we now get some nice di...

This gets rid of both the runtime overhead associated with dynamic_cast
and the need to explicit...

Instead tell CMake that it's a generated source file so that it accepts
their absence at generat...

This will cause the output to be displayed while cucumber executes our
features instead of when ...

Apply the same fix for cucumber as we applied for GCC/Clang to get
colored terminal output throu...

This fixes #153.

Because MSVC's value for __cplusplus doesn't even claim C++98 support.
That's actually good, bec...

This prevents nasty surprises where we might try to convert to a const
type, which would fail in...

On C++11 (and beyond), support extracting of the regex' captures to the
arguments of the step fu...

This is intended to provide a replacement for the use of the REGEX_PARAM
macro and to pass the p...

This allows one to bind the WireServer to localhost, which fixes an
annoying nag dialog shown by...

This fixes #156 by not depending on indirect inclusion and additionally
uses iswprint from the n...

Because I've seen this code to often, wondering why I shouldn't just
delete this line. Turns out...

I.e. inside FindGMock.cmake we're using find_dependency() from the
CMakeFindDependencyMacro modu...

I.e. running all cases in a single process, without respawning for every
test case, is now possi...

Also for maintainers with regards to release tagging.

HTTPS is the protocol least likely to be blocked by a corporate
firewall.

I.e. otherwise we don't ever validate the proper execution of SetUp()
and TearDown().

Except for Boost.Test, that's still optional.

Because we cannot build without that version, 1.40 isn't enough. 1.42 is
required for Boost.Test...

Closes #119

To be really sure we won't ever have a dangling pointer there.

Pass by reference instead of by pointer, where possible. Do this because
references must always ...

This makes it easier (with a simple grep '\bnew\b') to find the pointers
that really aren't trac...

This still doesn't suffer from the "static init order fiasco" because
the order of function-scop...

Fixes parallel builds (including Ninja)

Because after making fully parallel builds possible we'd better start
using them to keep testing...

This way other properties such as the list of libraries to link to
transitively and what header ...

This gives CMake the information it needs to figure out the dependencies
_properly_ by itself. A...

This prevents attempting to include things from
/usr/src/gmock/include/gmock/gmock.h, which in U...

Because _DIRS is an output variable that doesn't (yet) exist.

This pull request fixes the cmake files for building with MinGW under
windows. Further it modifi...