fixing interoperability with native child_process. also added tests

[jaredly]

calling waitpid when it was one of node's processes caused failure. This fixes it.

Should we perhaps use another map? Is there any case where SIGCHLD would be called multiple times for a single process, such that we would set the pidMap entry, and then not pay attention to successive events?

[niallo]

Thanks. Yeah in theory SIGCHLD could be called multiple times for a single child, although it's probably unlikely (child process would have to stop and then resume). Another map would work, or I think we could also set the SA_NOCLDSTOP bit in sa_flags to only receive SIGCHLD when the child exits, not when it stops.

[jaredly]

although we only do the check if it's an exit. right? isn't that what res
!= 0 does?

On Sat, Aug 3, 2013 at 1:18 AM, niallo [email protected] wrote:

Thanks. Yeah in theory SIGCHLD could be called multiple times for a
single child, although it's probably unlikely (child process would have to
stop and then resume). Another map would work, or I think we could also set
the SA_NOCLDSTOP bit in sa_flags to only receive SIGCHLD when the child
exits, not when it stops.

—
Reply to this email directly or view it on GitHubhttps://github.com//pull/1#issuecomment-22051092
.