Improved resource limits handling in the kqueue autoreloader.

    import resource

    NOFILES_SOFT, NOFILES_HARD = resource.getrlimit(resource.RLIMIT_NOFILE)

except AttributeError:

    import subprocess

    command = ["sysctl", "-n", "kern.maxfilesperproc"]

    NOFILES_KERN = int(subprocess.check_output(command).strip())

except (AttributeError, OSError):

    USE_KQUEUE = False

RUN_RELOADER = True

    Checks for changed code using kqueue. After being called

    it blocks until a change event has been fired.

    """

    # Maximum number of open file descriptors is typically too low (256).

    # We must increase the maximum number of open file descriptors because

    # kqueue requires one file descriptor per monitored file and default

    # resource limits are too low.

    #

    # In fact there are two limits:

    # - kernel limit: `sysctl kern.maxfilesperproc` -> 10240 on OS X.9

    # - resource limit: `launchctl limit maxfiles` -> 256 on OS X.9

    #

    # The latter can be changed with Python's resource module. However, it

    # cannot exceed the former. Suprisingly, getrlimit(3) -- used by both

    # launchctl and the resource module -- reports no "hard limit", even

    # though the kernel sets one.

    filenames = list(gen_filenames())

    resource.setrlimit(resource.RLIMIT_NOFILE,

                       (NOFILES_SOFT + len(filenames), NOFILES_HARD))

    # If project is too large or kernel limits are too tight, use polling.

    if len(filenames) > NOFILES_KERN:

        return code_changed()

    # Add the number of file descriptors we're going to use to the current

    # resource limit, while staying within the kernel limit.

    nofiles_target = min(len(filenames) + NOFILES_SOFT, NOFILES_KERN)

    resource.setrlimit(resource.RLIMIT_NOFILE, (nofiles_target, NOFILES_HARD))

    kqueue = select.kqueue()

    fds = [open(filename) for filename in filenames]