Here is an easily fixable semi-bug in macostools:
When two threads are creating the same directory structure, one 
of them will fail as soon as the other thread made a directory.
This is trivially fixable by ignoring the error, but I'm not sure if it is 
worth it, like this:

def mkdirs(dst):
    """Make directories leading to 'dst' if they don't exist yet"""
    if dst == '' or os.path.exists(dst):
        return
    head, tail = os.path.split(dst)
    if os.sep == ':' and not ':' in head:
        head = head + ':'
    mkdirs(head)
    try: os.mkdir(dst, 0777)
    except OSError, err:
	if err.errno==17: #file exists
		#someone else has created the directory in the 
meantime. That's fine with me!
		pass
	else: raise
- Jurjen

Logged In: YES 
user_id=92689

Hm, macostools.mkdirs() also isn't portable: it only works on OS9 
MacPython. Why don't you use os.makedirs() instead? (Which appears to 
have the same threads issue, though.)

Logged In: YES 
user_id=92689

Oops, disregard that portability remark, it does work correctly in non-OS9 
Python as well. I _think_ macostools.mkdirs() predates os.makedirs(), it's 
probably best considered deprecated.

Logged In: YES 
user_id=139309

Most Python functions aren't intended to be thread-safe, this is what locks are 
for. I don't think this should be fixed, not even in os.makedirs.

Fixed this issue in r68161 by porting the solution for this in os.mkdirs 
to macostools.