- To avoid getting stuck in a loop, repeated EBADF failures
- are 'handled' through barfing.
+ Note: assume here that threads becoming unblocked
+ will try to read/write the file descriptor before trying
+ to issue a threadWaitRead/threadWaitWrite again (==> an
+ IOError will result for the thread that's got the bad
+ file descriptor.) Hence, there's no danger of a bad
+ file descriptor being repeatedly select()'ed on, so
+ the RTS won't loop.