On Tue, 30 Jan 2024 10:47:22 GMT, Sebastian Lövdahl <d...@openjdk.org> wrote:
> 8307977: jcmd and jstack broken for target processes running with elevated > capabilities Hi, Yes makes sense, this seems like an oversight that we were not consistent with the path. Does CAP_NET_BIND_SERVICE cause any issues for createAttachFile(int pid, int ns_pid) where it creates the .attach file in the current directory - it starts by trying "/proc/" + pid + "/cwd/" + ".attach_pid" + ns_pid, regardless of ns_pid. I'm curious if that always fails in the situation that causes the issue in this bug. If so looks like it would catch an IOException and then use findTargetProcessTmpDirectory, but wonder if we should predict it go straight there. Thanks! ------------- PR Comment: https://git.openjdk.org/jdk/pull/17628#issuecomment-1930386568