> I'd like to commit the attached patch to the 3.4 branch. This was > discovered by the unit test which failed when an empty matrix element > didn't have a value of 0.0. As it turned out, I was initializing the > higher 4 bytes via 32-bit pointer but the lower 4 bytes were left > un-initialized, which caused the double value to become non-zero. > Apparently this doesn't happen when building a 64-bit build.
This indeed avoids the assertion failure in the unit test. I first tried a change to initialising the fVal and not pS only in the no-argument constructor, but that didn't help. Kohei's additional changes to the copy constructor and operator= did help, great! Pushed. --tml _______________________________________________ LibreOffice mailing list [email protected] http://lists.freedesktop.org/mailman/listinfo/libreoffice
