https://bugs.kde.org/show_bug.cgi?id=386945
--- Comment #5 from Markus Trippelsdorf <octopl...@yandex.com> --- (In reply to Julian Seward from comment #2) > This is caused by insufficiently precise definedness analysis for > (1) count leading zero operations (Iop_Clz32, Iop_Clz64) > (2) PPC integer compares (Iop_CmpORD family) > (3) integer subtract > > Here's an initial, hacky patch that fixes (1) and (2). Along with > use of the flag --expensive-definedness-checks=yes, which fixes (3), > it makes the test case run quiet. > > Can you give it a try on something larger? Unfortunately it does't work yet. I still get "Invalid read of size 4" and "Conditional jump or move depends on uninitialised value(s)" on many strcmp() invokations. Plus the patch corrupts memory and gcc ICEs now. (Testcase is running g++ under valgrind when compiling a small C++ program.) -- You are receiving this mail because: You are watching all bug changes.
