https://bugs.kde.org/show_bug.cgi?id=365327
--- Comment #15 from Austin English <austinengl...@gmail.com> --- (In reply to Louis Brunner from comment #12) > Created attachment 104964 [details] > macOS Sierra incomplete support > > I have been working on a patch to support macOS Sierra. > At the moment, it works for a variety of programs (python, GIMP, most basic > commands like ls, mkdir) but it is still incomplete (warnings in dyld, > libsystem_kernel... crash for some GUI programs). > > The patch adds a few required syscalls with placeholder implementations > (faccessat, fstatat64, csrctl, getentropy and ulock_wake) and the new way of > loading dylib (placing them at the end of the currently loaded segments). > The second change means we need to know where the last segment was loaded, > which means carrying around one more pointer on pointer in every function > (which already have 6-9 arguments), that's why I created a structure > (load_info_t) to store all this information and easily carry it around. > > It also adds one assert in is_in_syscall in > coregrind/m_syswrap/syswrap-main.c to match the other syscall related > functions in the same file. I had a difficult to diagnose crash in this > function because it didn't check for the existence of the syscall table. > > Tell me if you need any change I can confirm that the situation is improved on 10.12 with this patch. E.g., valgrind will at least attempt things now. I noticed that `./vg-in-place ls -al` crashes valgrind, though `./vg-in-place true` works fine (does show some uninitialized memory in OSX though). A basic sanity check on 10.11 also worked for me. -- You are receiving this mail because: You are watching all bug changes.
