DesktopLOKTest::testRedlineWriter is failing

[Stephan Bergmann]

On 11/10/2017 10:17 AM, Tomáš Chvátal wrote:
> 2017-11-06 9:58 GMT+01:00 Stephan Bergmann <sbergman at redhat.com 
> <mailto:sbergman at redhat.com>>:
> 
>     Looking at the surrounding code
> 
>              // Get redline info.
>              boost::property_tree::ptree aTree;
>              char* pJSON =
>         pDocument->m_pDocumentClass->getCommandValues(pDocument,
>         ".uno:AcceptTrackedChanges");
>              std::stringstream aStream(pJSON);
>              free(pJSON);
>              CPPUNIT_ASSERT(!aStream.str().empty());
>              boost::property_tree::read_json(aStream, aTree);
>              // Make sure that pressing a key creates exactly one redline.
>              CPPUNIT_ASSERT_EQUAL(static_cast<size_t>(1),
>         aTree.get_child("redlines").size());
> 
> 
>     what's the binary content of that "char* pJSON" stream?  If that's
>     still intact, then it's apparently boost::property_tree::read_json
>     that's breaking things.  And if that's already garbled, then it's
>     apparently the producer side of that
>     pDocument->m_pDocumentClass->getCommandValues (whatever that is)
>     that is breaking it and needs further investigation.
> 
> I checked the read_json function and it was indeed rewritten in between 
> the boost versions on the distro (and there were some comments about 
> unicode support). So I had updated the boost function (basically to use 
> latest boost code) in my testing repo and it still fails:
> 
> [ 6427s] /home/abuild/rpmbuild/BUILD/libreoffice-5.4.2.2/desktop/qa/desktop_lib/test_desktop_lib.cxx:1660:DesktopLOKTest::testRedlineWriter
> [ 6427s] equality assertion failed
> [ 6427s] - Expected: Insert “t”
> [ 6427s] - Actual  : Insert “tâ€
> 
> 
>   FWIW I enabled it on the OBS so the full build log is tracked here:
> https://build.opensuse.org/package/live_build_log/LibreOffice:Factory/libreoffice/openSUSE_Leap_42.3/x86_64
> 
> Any idea how to check the getCommandValues behaviour if it is the proper 
> one?

so, again, "what's the binary content of that "char* pJSON" stream?" 
can't you just dump it in gdb?