Patch by David CARLIER
this should not be needed after r328077
What about this?
Why zeroes? Can you get real values?
I commented this commit here
I thought I did this one ...
ucontext not supported here.
Nice works better under FreeBSD should do the job under OpenBSD.
newfstatat should return int, so no _ptr
same for stat
same for unlinkat, it returns int, on both
Sorry, that was a typo.
Sorry, but I only asked, because I am not sure why you are making this change. According sanitizer_common/sanitizer_syscall_generic.inc
internal_syscall_ptr for pointer and (s)size_t
Looking on documentation all of them return int and internal_syscall supposed to be used. What am I missing?
https://man.openbsd.org/sigaction.2
On OpenBSD, ucontext_t is an alias for the sigcontext structure defined in <signal.h>.
Seems under Linux it s the same in fact https://github.com/llvm-mirror/compiler-rt/blob/b4e600115ad4c50192095c225aaed3ae9ca99ec1/lib/sanitizer_common/sanitizer_syscall_generic.inc#L53 but I personally do not mind reverting them I think it was kamel who asked me to update in the first place.
it's the same, but if we introduced internal_syscall_ptr and we specified that it's for pointers and size, we should follow that convention.
Good enough for me.
I think it should be internal_syscall() as it's int everywhere, why do you need ptr? And this is not OpenBSD branch at all
On linux readlinkat is int, on OpenBSD it's size. So it should be two different branches
it's int on OpenBSD as well https://github.com/openbsd/src/blob/master/sys/sys/syscall.h
Could you please comment about your plans here?
All good points ... Did not know about sigcontext
Adding sigcontext support
What do you think about this one?
Sorry I missed this one good point.