* handle interrupted reads from the semaphore pipe in Darwin (should be
applied to *BSD and Linux as well)
+ lots of extra debugging code in Darwin tthread.inc if DEBUG_MT defined
git-svn-id: trunk@1416 -
suspended thread, it was possible that resume was executed before
that thread had completed its initialisation in BeginThread ->
FInitialSuspended was set to false in resume and nevertheless a
semafore was posted
* second problem fixed: set FSuspended to false before waking up the
thread, so that it doesn't get FSuspended = true right after waking
up. This should be done atomically to be completely correct though.
