openoffice-dev mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From Damjan Jovanovic <dam...@apache.org>
Subject Re: cppunit -> Google Test migration and old failing tests
Date Mon, 31 Aug 2015 06:52:34 GMT
On Mon, Aug 31, 2015 at 5:54 AM, Damjan Jovanovic <damjan@apache.org> wrote:
> On Sun, Aug 30, 2015 at 11:16 PM, Kay Schenk <kay.schenk@gmail.com> wrote:
>>
>>
>> On 08/29/2015 12:37 AM, Damjan Jovanovic wrote:
>>> On Fri, Aug 28, 2015 at 6:07 PM, Kay Schenk <kay.schenk@gmail.com> wrote:
>>>>
>>>> On 08/27/2015 09:05 PM, Damjan Jovanovic wrote:
>>>>> Hi
>>>>>
>>>>> I am in the process of migrating our unit tests from cppunit to Google
>>>>> Test. However AOO doesn't build with cppunit and hasn't been routinely
>>>>> built with cppunit for a while, which means our unit tests are in a
>>>>> state of neglect, and unsurprisingly, there are many failures both
>>>>> compiling and running our unit tests.
>>>>>
>>>>> Ideally we should investigate why and fix the tests. But the APIs
>>>>> being tested are complex and unfamiliar to me (eg. SVG parsing), and
>>>>> would take very long to investigate properly.
>>>>>
>>>>> I could commit changes that will just get the tests to compile, then
>>>>> fail during testing and stop the build, thus forcing others to fix
>>>>> them quickly :-), but I don't imagine that will go down well. So I am
>>>>> taking this approach instead:
>>>>>
>>>>> // FIXME:
>>>>> #define RUN_OLD_FAILING_TESTS 0
>>>>>
>>>>> #if RUN_OLD_FAILING_TESTS
>>>>> broken_test();
>>>>> #endif
>>>>>
>>>>> Also I am making unit tests run on every build. This way at least some
>>>>> unit tests will be run, and any future regressions to tests can be
>>>>> caught immediately, while the broken tests can be fixed gradually.
>>>>>
>>>>> Everyone happy?
>>>>
>>>> Well pretty much. :)
>>>>
>>>> I've been watching your commits. Thank you for taking on this
>>>> challenging task.
>>>
>>> Thank you.
>>>
>>>> OK, just to be clear. It looks like you're converting the cppunit calls
>>>> to Google Test api calls. But, what you're saying is the actual use of
>>>> the Google test routines needs additional modification to work
>>>> correctly, right?
>>>
>>> Yes that's what I am doing.
>>>
>>> No, the C++ conversions are very easy (feel free to help ;-)):
>>> #include "cppunit..."   =>   #include "gtest/gtest.h"
>>> class X : public CppUnit::TestFixture  =>  class X : public ::testing:Test
>>> CPPUNIT_ASSERT_MESSAGE(msg, condition)   =>   ASSERT_TRUE(condition) <<
msg
>>> CPPUNIT_ASSERT_EQUAL(c1, c2)    =>   ASSERT_EQ(c1, c2)
>>> CPPUNIT_FALSE(msg)   =>   FAIL() << msg
>>> private:  =>  protected:
>>> test methods move outside of class declaration and become
>>> TEST_F(className, methodName) instead
>>> CPPUNIT_TEST...() registrations disappear
>>>
>>> but the problem is that tests themselves are wrong no matter what the
>>> testing library. For example:
>>> basegfx::tools::importFromSvgD( aPoly, aSvg );
>>> won't compile, as importfromSvgD() requires 4 parameters now instead
>>> of just 2 (as I explained in an earlier email, this was caused by
>>> commit 1536730 on 2013-10-29 by alg).
>>>
>>> Damjan
>>
>> A quick question on these changes. Do these require a reconfigure to
>> work correctly? I ran into a problem building with r1698208 so this is
>> why I ask.
>>
>
> I see, main/codemaker's tests don't build unless AOO is already built;
> it's probably one of those tests that needs OOO_SUBSEQUENT_TESTS.
>
> r1698208 builds for me with this patch (ie. delete the last line of
> text in main/codemaker/prj/build.lst):
>
> Index: main/codemaker/prj/build.lst
> ===================================================================
> --- main/codemaker/prj/build.lst    (revision 1700184)
> +++ main/codemaker/prj/build.lst    (working copy)
> @@ -7,4 +7,3 @@
>  cm    codemaker\source\cppumaker                nmake    -    all
> cm_cppumaker     cm_codemaker cm_cpp cm_inc NULL
>  cm    codemaker\source\commonjava                nmake    -    all
> cm_java cm_inc NULL
>  cm    codemaker\source\javamaker                nmake    -    all
> cm_javamaker cm_codemaker cm_java cm_inc NULL
> -cm    codemaker\test\cppumaker                nmake    -    all
> cm_cppumaker_test cm_cppumaker cm_codemaker cm_cpp cm_inc NULL
>
> However in the latest SVN I then get another build error in main/oovbaapi:
> Compiling: Globals.idl
> /tmp/idli_wyec6x: line 32: file 'com/sun/star/table/XCellRange.idl' not found
> AOO/main/solver/420/unxfbsdx/bin/idlc: preprocessing file
> /AOO/main/oovbaapi/ooo/vba/excel/Globals.idl failed
> dmake:  Error code 1, while making '../../../unxfbsdx/ucr/excel.db'
>
> which isn't in a module I changed. I am bisection testing between
> r1698208 and r1700184 to see what broke that.

I fixed several problems and am busy rebuilding. Please check if the
latest SVN works for you. If not, please post the build log.

---------------------------------------------------------------------
To unsubscribe, e-mail: dev-unsubscribe@openoffice.apache.org
For additional commands, e-mail: dev-help@openoffice.apache.org


Mime
View raw message