…o avoid to release the semaphore twice. Some deferred check to speed up things

…tive(pipe_fd) returns false. Use SetBlockingMode() to set O_NONBLOCK for pipes instead of SetMode().

Fix problems with pipe and O_NONBLOCK.

…or old and new name

…__get_default_file is now public but needs to be declared as extern if someone wants to use it (or include dos.h that can come with side effects)

Fixed Spawnv. Added waitpid/wait functions

Reverse the order of SBM_(NON_)BLOCKING. (Second attempt.)

…no().

Added ERROR_WOULD_BLOCK == EWOULDBLOCK to __translate_io_error_to_err…

… test if tty functions are working correctly (and it seems not..)

Added handling of O_NONBLOCK fds for select/poll.

Poll mode reported avaibale data,even that isn't the case

Removed old function that was just a wrapper of ParentOfFH
Fixed a problem on select_signal.c was causing some problems on PIPEs

…-amiga path translation utility functions.

Added private function __search_expand_command_path()

…o because it strips away the constructors..)

…e called and everything is safe at that point

…kfile.. needs to fix it

…on sigaction was not reentrant

… time. funlock file now check if lock is from correct task

…y to use an #id as pthread

…ation, because it can break the resulting program in cases, where such an object exists on the system path but not in the current dir (stat(EXISTS) will return true in such cases, where the correct answer is false). 2) Comment out/suggest removal of the snip, that tags on the current path to an immediate (no-path) object in unix->amiga path translation, because of a parallel and very opposite case : If the program is trying to execute a command on the system search path, this snip will tag on code, that makes it impossible for the call (for instance popen(command) to find the command).