Test name: Test::testAutoFill assertion failed

Alex McMurchy mcmurchy1917-libreoffice at yahoo.co.uk
Fri Oct 23 22:13:26 PDT 2015


Hi All

Some more information on this problem.

It occurs on Slackware 14.1 and Current - the compiler is gcc versions 
4.82 and 4.93 respectively.

It does NOT occur in Lubuntu 14:04- the compiler is gcc version 4.84


The results of a bisect are -

> 8e4dc1d760d85e09bbc3f3bbb5b8be2947db1b63 is the first bad commit
> commit 8e4dc1d760d85e09bbc3f3bbb5b8be2947db1b63
> Author: Noel Grandin <noel at peralex.com>
> Date:   Fri Aug 22 15:15:17 2014 +0200
>
>     create type-safe bitfield for sc insert/delete flags
>
>     The most important part of the change is in sc/inc/global.hxx
>
>     It creates a type-safe struct that prevents the accidental interaction
>     between regular integer types and the flags struct.
>     It also provides utility methods that make combining and testing the
>     flags type-safe.
>
>     Change-Id: Ibc5b20058b1655df913490682b679afd1297b36d
>     Reviewed-on: https://gerrit.libreoffice.org/11071
>     Reviewed-by: Eike Rathke <erack at redhat.com>
>     Tested-by: Eike Rathke <erack at redhat.com>
>
> :040000 040000 f9f6a67e9269712bb1997ee6ba9fb49d21e4455e 
> 494c585ff088b7980af64f28e556a58b2afcc35b M      sc

I've had a look at the contents of the commit and it does look as though 
it touches the area that's causing me problems. The commit does look 
fairly generic so why one distro should fail and another pass does look 
a bit bizarre.

I suppose the next stage is to pull out gdb from the toolkit and try and 
ascertain where the paths of Lubuntu and Slackware diverge.

Or, has anyone got any other suggestions on how to tackle this problem?

I've not raised a bug report yet.

Alex



On 01/09/15 18:22, Alex McMurchy wrote:
> Hi All
>
> When compiling LibreOffice I get a failed Unit test. If I hash out the 
> failing CPPUNIT_ASSERT I get no more errors and Libreoffice compiles 
> successfully.
>
> I suspect the problem is environmental, missing package or something, 
> as I've managed to compile without intervention on another distro.
>
> Any suggestions on how I can resolve this problem, without hashing out 
> the code or skipping the test, will be appreciated.
>
> I've pulled from git and git branch says I'm on
>
> * libreoffice-5-0-0
>   master
>
> I've compiled with
>
> --without-gssapi --without-krb5 --without-system-npapi-headers
>
>
> The error is -
>
>> [build CXX] sc/qa/unit/ucalc.cxx
>> [build CXX] sc/qa/unit/ucalc_column.cxx
>> [build CXX] sc/qa/unit/ucalc_formula.cxx
>> [build CXX] sc/qa/unit/ucalc_pivottable.cxx
>> [build CXX] sc/qa/unit/ucalc_sharedformula.cxx
>> [build CXX] sc/qa/unit/ucalc_sort.cxx
>> [build DEP] LNK:CppunitTest/libtest_sc_ucalc.so
>> [build LNK] CppunitTest/libtest_sc_ucalc.so
>> [build CUT] sc_ucalc
>> /home/master/git/libreoffice/sc/qa/unit/ucalc.cxx:4540:Test::testAutoFill 
>>
>> assertion failed
>> - Expression: m_pDoc->GetCellType(ScAddress(0,i,0)) == CELLTYPE_NONE
>>
>> ucalc.cxx:4540:Assertion
>> Test name: Test::testAutoFill
>> assertion failed
>> - Expression: m_pDoc->GetCellType(ScAddress(0,i,0)) == CELLTYPE_NONE
>>
>> Failures !!!
>> Run: 205   Failure total: 1   Failures: 1   Errors: 0
>>
>> Error: a unit test failed, please do one of:
>>
>> export DEBUGCPPUNIT=TRUE            # for exception catching
>> export CPPUNITTRACE="gdb --args"    # for interactive debugging on Linux
>> export VALGRIND=memcheck            # for memory checking
>>
>> and retry using: make CppunitTest_sc_ucalc
>>
>> make[1]: *** 
>> [/home/master/git/libreoffice/workdir/CppunitTest/sc_ucalc.test] Error 1
>> make: *** [build] Error 2
>> Compilation of libreoffice failed with 2 at Sun 30 Aug 16:01:04 BST 2015
>
> If I hash out the problem area like this -
>
>>  4536      pUndoMgr->Undo();
>>   4537
>>   4538  /*    CPPUNIT_ASSERT_EQUAL(1.0, 
>> m_pDoc->GetValue(ScAddress(0,0,0)));
>>   4539      for (SCROW i = 1; i <= 5; ++i)
>>   4540 CPPUNIT_ASSERT(m_pDoc->GetCellType(ScAddress(0,i,0)) == 
>> CELLTYPE_NONE); */
>
> The test no longer fails and LibreOffice goes on to compile successfully.
>
> I've done a quick debug on it and the failure occurs on the first 
> iteration of the loop when i=1
>
> Alex
>
>
>
> _______________________________________________
> LibreOffice mailing list
> LibreOffice at lists.freedesktop.org
> http://lists.freedesktop.org/mailman/listinfo/libreoffice



More information about the LibreOffice mailing list