[PATCH v2 2/3] Add XInput 2.x integration test framework

[Chase Douglas]

On 05/15/2012 04:26 PM, Peter Hutterer wrote:
> On Tue, May 15, 2012 at 11:05:26AM -0700, Chase Douglas wrote:
>> Signed-off-by: Chase Douglas <chase.douglas at canonical.com>
>> ---
>> Changes since v1:
>> * Split commit into XInput 2.x integration test framework and XIQueryPointer
>>   test
>>
>>  configure.ac                 |   14 +++
>>  test/integration/.gitignore  |    1 +
>>  test/integration/Makefile.am |   24 ++++++
>>  test/integration/xi2.cpp     |  194 ++++++++++++++++++++++++++++++++++++++++++
>>  4 files changed, 233 insertions(+)
>>  create mode 100644 test/integration/.gitignore
>>  create mode 100644 test/integration/xi2.cpp
>>
>> diff --git a/configure.ac b/configure.ac
>> index fe350c9..bc9f46f 100644
>> --- a/configure.ac
>> +++ b/configure.ac
>> @@ -2141,6 +2141,20 @@ if [test "x$XORG" = xyes && test "x$enable_integration_tests" != xno]; then
>>      fi
>>  fi
>>  
>> +PKG_CHECK_MODULES(TEST_X11, x11, have_test_x11=yes, have_test_x11=no)
>> +PKG_CHECK_MODULES(TEST_XINPUT, [xi >= 1.6], have_test_xinput=yes, have_test_xinput=no)
>> +
>> +if [test "x$have_test_x11" != xyes]; then
>> +    AC_MSG_NOTICE([libX11 not available, skipping input tests])
>> +elif [test "x$have_test_xinput" != xyes]; then
>> +    AC_MSG_NOTICE([libXi 1.6 not available, skipping input tests])
>> +elif [test "x$have_xorg_gtest_evemu" != xyes]; then
>> +    AC_MSG_NOTICE([uTouch-Evemu not available, skipping input tests])
>> +else
>> +    enable_input_tests=yes
>> +fi
>> +AM_CONDITIONAL(ENABLE_XORG_GTEST_INPUT_TESTS, [test "x$enable_input_tests" = xyes])
>> +
>>  dnl and the rest of these are generic, so they're in config.h
>>  dnl 
>>  dnl though, thanks to the passing of some significant amount of time, the
>> diff --git a/test/integration/.gitignore b/test/integration/.gitignore
>> new file mode 100644
>> index 0000000..ab86ebf
>> --- /dev/null
>> +++ b/test/integration/.gitignore
>> @@ -0,0 +1 @@
>> +gtest-tests
>> diff --git a/test/integration/Makefile.am b/test/integration/Makefile.am
>> index e70d642..3b7c858 100644
>> --- a/test/integration/Makefile.am
>> +++ b/test/integration/Makefile.am
>> @@ -1,4 +1,28 @@
>> +TESTS =
>> +
>>  if ENABLE_XORG_GTEST_TESTS
>>  include $(top_srcdir)/test/integration/Makefile-xorg-gtest.am
>>  check_LIBRARIES = $(XORG_GTEST_BUILD_LIBS)
>> +
>> +if ENABLE_XORG_GTEST_INPUT_TESTS
>> +TESTS += gtest-tests
>>  endif
>> +endif
>> +
>> +check_PROGRAMS = $(TESTS)
> 
> noinst_PROGRAMS is better here.

Ok.

>> +
>> +AM_CPPFLAGS = \
>> +	-DDEFAULT_XORG_SERVER=\"$(abs_top_builddir)/hw/xfree86/Xorg\" \
>> +	-DTEST_ROOT_DIR=\"$(abs_top_srcdir)/test/integration/\" \
>> +	$(GTEST_CPPFLAGS) \
>> +	$(XORG_GTEST_CPPFLAGS)
>> +
>> +AM_CXXFLAGS = $(GTEST_CXXFLAGS) $(XORG_GTEST_CXXFLAGS)
>> +
>> +gtest_tests_SOURCES = xi2.cpp
>> +gtest_tests_LDADD = \
>> +	$(TEST_XINPUT_LIBS) \
>> +	$(TEST_X11_LIBS) \
>> +	$(XORG_GTEST_LIBS) \
>> +	$(EVEMU_LIBS) \
>> +	$(XORG_GTEST_MAIN_LIBS)
>> diff --git a/test/integration/xi2.cpp b/test/integration/xi2.cpp
>> new file mode 100644
>> index 0000000..68974a9
>> --- /dev/null
>> +++ b/test/integration/xi2.cpp
>> @@ -0,0 +1,194 @@
>> +#include <stdexcept>
>> +
>> +#include <xorg/gtest/xorg-gtest.h>
>> +
>> +#include <X11/extensions/XInput2.h>
>> +
>> +namespace {
>> +
>> +/**
>> + * Wait for an event on the X connection.
>> + *
>> + * param [in] display The X display connection
>> + * param [in] timeout The timeout in milliseconds
> 
> shouldn't these be @param?

Yep, thanks for catching it.

> Reviewed-by: Peter Hutterer <peter.hutterer at who-t.net> otherwise
> 
> one question that remains though: my impression was that xorg-gtests would
> go into the xorg-gtest repo but these are for the xserver repo. What's the
> plan here?

These tests are testing the X server itself. I believe that having the
tests in the upstream project repo ensures they are more likely to be
run during development. I don't really see a reason for why they should
live somewhere else.

> wait_for_event on a Display* seems generic enough that I don't quite
> understand why this isn't part of the xorg-gtest repo.

It likely will migrate there. However, I'm not 100% confident of its
structure yet, so I don't want to commit to it in the xorg-gtest API/ABI
quite yet. I would like to see more tests written using it first.

-- Chase