Failpoints are often injected at system call sites (send/recv). Typically the error-occurred branch of the nearby code for system calls will examine 'errno' and respond to that value in some way (often by capturing it in an exception). If, however, the error-occurred branch is taken because the failpoint is enabled, rather than because the syscall failed, then the value of errno is indeterminate, and therefore the path through the error handling logic cannot be controlled.

The failpoints mechanism should offer a facility to associate an errno value with an activated failpoint, so that when the failpoint triggers errno is set to the stored value. This would make it possible to control the behavior of the error-occurred branch when the failpoint is active.