Back

I nailed the #libreoffice bug I mentioned the other day; and #rr definitely helped me - in a code base I didn't know that well, finding what set a value was really nice and led me (together with some std::cerr debug and a bit of normal gdb) to the buggy function. But also, I could reverse-continue from a working path, and find the equivalent code that path used - net result was sharing the working code to the bad path; one dead bug, 40 less lines, all happy.