def _delay_for_pipe(self, env=None, delay=None):
# This delay is sometimes needed because the temporary stdin file that
# is being written (the pipe) may not have even hits its first flush()
# call by the time the spawned process starts up and determines there
# is nothing in the file. The spawn can thus exit, without doing any
# real work. Consider the case of piping something into grep:
#
# $ ps aux | grep root
#
# grep will exit on EOF and so there is a race between the buffersize
# and flushing the temporary file and grep. However, this race is not
# always meaningful. Pagers, for example, update when the file is written
# to. So what is important is that we start the spawned process ASAP:
#
# $ ps aux | less
#
# So there is a push-and-pull between the the competing objectives of
# not blocking and letting the spawned process have enough to work with
# such that it doesn't exit prematurely. Unfortunately, there is no
# way to know a priori how big the file is, how long the spawned process
# will run for, etc. Thus as user-definable delay let's the user
# find something that works for them.
if delay is None:
delay = (env or os.environ).get('TEEPTY_PIPE_DELAY', -1.0)
delay = float(delay)
if 0.0 < delay:
time.sleep(delay)
评论列表
文章目录
