IRC exchange:

<tml> reverting 16bfddca makes the CppunitTest_sc_subsequent_filters_test not
crash for me (Linux, with the AMDAPPSDK-2.9-1 OpenCL implementation, but no
AND hardware, just an (Intel) CPU), but then it instead fails the
ScFiltersTest::testSharedFormulaXLS

<tml> the problem (in the non-reverted case) is that for some reason the
pDVR->GetArrayLength() on line 131 returns a ridiculously large value in
sc/source/core/opencl/formulagroupcl.cxx

<tml> that seems to cause the alllocated buffer to actually be of zero size,
according to valgrind (even if no OpenCL call returns failure...)

<moggi> let me have aquick look

<moggi> just comment out that line for now and I'll have a look at it during
the weekend

Change-Id: I5f9499ba1576cfa6d0945ecb19260a4b89c83e72