Re: Tests failures trying to package LO in Fedora

[libreoffice@xxxxxxxxxxxxxxxxxxxxxxxx]

Il 07/08/23 11:55, Eike Rathke - erack at redhat.com ha scritto:
> Hi,
>
> On Monday, 2023-08-07 06:12:39 +0000, libreoffice@xxxxxxxxxxxxxxxxxxxxxxxx wrote:
>
>> /builddir/build/BUILD/libreoffice-7.6.0.2/sc/qa/unit/ucalc_formula.cxx:1650:testFormulaRefUpdate::TestBody
>> equality assertion failed
>> - Expected: =SUM(D2:E3)
>> - Actual  : =SUM(B2:E3)
>> - Wrong formula in A6.
> If that test fails, something is fundamentally broken. The columns
> inserting code just above on line 1647 does not work or does not adjust
> formula expression references
>
>      // Insert columns B:C to shift only the value range.
>      m_pDoc->InsertCol(ScRange(1,0,0,2,m_pDoc->MaxRow(),0));
>
> so you'd have to debug step into that and see where it fails. Use the
>
> make CppunitTest_sc_ucalc_formula CPPUNITTRACE="gdb --args"
>
> command given by the build output, set a breakpoint on
> sc/qa/unit/ucalc_formula.cxx:1647 and run.

I never used gdb... I used the make command above, then inserted a 
breakpoint with 'break sc/qa/unit/ucalc_formula.cxx:1647', then issued 
'run'... and I end with the gdb prompt without any error. What else?


>
> However, the build seems to have a more general problem, the build.log
> output is full with plenty of "exceeds maximum object size" like (first
> occurrence)
>
> | /builddir/build/BUILD/libreoffice-7.6.0.2/cppuhelper/source/component_context.cxx:571:87: warning: argument 1 value '18446744073709551615' exceeds maximum object size 9223372036854775807 [-Walloc-size-larger-than=]
> |   571 |     std::unique_ptr<ContextEntry_Init[]> mapped_entries(new ContextEntry_Init[nEntries]);
> |       |                                                                                       ^
> | /usr/include/c++/13/new:128:26: note: in a call to allocation function 'operator new []' declared here
> |   128 | _GLIBCXX_NODISCARD void* operator new[](std::size_t) _GLIBCXX_THROW (std::bad_alloc)
> |       |                          ^
>
> Those certainly don't look healthy..

There were a few of those also in 7.5.2 build with the same buildroot 
and system libraries, just a lot fewer. Any idea how to find the cause?

Thanks
Mattia