https://bugs.kde.org/show_bug.cgi?id=386945
--- Comment #20 from Julian Seward <jsew...@acm.org> --- (In reply to Aaron Sawdey from comment #19) > Yes, the branch on uninitialized data is intended Thanks for the clarification. Unfortunately, the analysis framework has the deeply wired-in assumption that every conditional branch instruction is "important" for the final outcome of the program. So there's no easy way (AFAIK, at least) to fix it from the Memcheck side. This isn't the first time I've seen this kind of thing -- a conditional branch on uninitialised data, followed by suitable fixups afterwards. I think gzip/zlib used to do this, but were subsequently modified so as to not do that. Despite some study of the problem I haven't come up with a way to solve it. How practical is it for you to change the generated code so it doesn't do that? There's some related stuff in the talk at https://fosdem.org/2018/schedule/event/debugging_tools_memcheck/. See in particular the second half of slide 13. -- You are receiving this mail because: You are watching all bug changes.
