-
Stephan Bergmann yazdı
This e.g. reduces the LSan-reported leakage during CppunitTest_sd_filters_test from SUMMARY: AddressSanitizer: 115048 byte(s) leaked in 4639 allocation(s). to SUMMARY: AddressSanitizer: 6544 byte(s) leaked in 118 allocation(s). The history is that this happened to (presumably accidentally) get fixed when switching from an SV_DECL_PTRARR to a boost::ptr_vector in b35980d9 "Move away from SV_DECL_PTRARR in CharAttribList" but then got "fixed" again with 46da0b68 "further sc/editeng unbreaking." That latter commit doesn't give any more detail than "this boost::ptr_vector idea is seriously broken," but I assume it did that change only via code inspection, to return to the previous (assumed good) status quo, rather than because that recently introduced deletion actually caused any problems there. I now see that this has also independently been found and fixed in AOO with <https://bz.apache.org/ooo/show_bug.cgi?id=120865> "Part of EditAttrib objects are not released after removing from Character attributes list" (the fix for which apparently got mis-classified for LO as "dd127fae prefer: e4e4b3d6," i.e., "Another SV_DECL_PTRARR now gone" about some unrelated XBaseParaPortionList). Change-Id: Iecee5a4c85bc0c0c36a6690d70f0a4a59dab964a
0f6f2134