On Wed, Feb 5, 2020 at 11:33 PM Manlio Perillo <manlio.peri...@gmail.com> wrote: > > Is there some overhead due to preemption?
Sure, but measurements show that performance is better with this approach, especially at scale. > About EINTR, I suspect that some programs will have problems. > As an example > https://github.com/golang/go/issues/22838 > > The issue was reported for MacOS, but technically every signal, including > SIGSTOP, can interrupt a syscall. Go installs all signal handlers with SA_RESTART set, so most cases will not see an EINTR error. And the Go standard library already checks for EINTR every place that it needs to. The issue you mention is not a problem on most systems; apparently on macOS SA_RESTART doesn't work for SIGTSTP. That said, as the release notes say, some programs that make direct system calls are going to see more EINTR errors than they did before. EINTR was already a possibility; it's just more common now. Ian -- You received this message because you are subscribed to the Google Groups "golang-nuts" group. To unsubscribe from this group and stop receiving emails from it, send an email to golang-nuts+unsubscr...@googlegroups.com. To view this discussion on the web visit https://groups.google.com/d/msgid/golang-nuts/CAOyqgcW66BwS4GQecwGLVHAQ8vXJTjc_OU_YSPQwfkq2ZqC7aA%40mail.gmail.com.
