I don't know if what I intend to do is even possible or reasonable, so I'm open for any suggestions.
Currently I have a script which starts n subprocesses of some.exe id which I regularly poll() to determine if they terminated and if so, with which errorlevel. The subprocesses are saved in a dict (key = subprocess, value = id). n instances of some.exe are kept running, each with an own id, until every id of a predefined list has been processed.
some.exe has no gui on it's onw, it writes the progress to the stdout, which I do not need. Now, for some reason sometimes, some.exe doesn't continue, as if it would wait - thus poll() never produces an errorlevel and done = proc.poll() is not None is never true. Which sooner or later leads to my dict of n procs, all being inactive and the overall progress is stuck.
If issued manually in a cmd, some.exe with an id, that shows this behaviour in the script - works perfectly fine.
Therefore my idea was to start a new cmd window from the script, which runs some.exe with the id, but I should still be able to poll() said exe.
Here's roughly what I have so far:
while id_list > 0:
if len(proc_dict) < n:
id = next_id()
proc = subprocess.Popen(["some.exe", id], stdout=PIPE)
proc.poll()
proc_dict[proc] = id
else:
done_procs = []
for proc in proc_dict.keys():
done = proc.poll() is not None
if done:
print("returncode: "+proc.returncode)
done_procs.append(proc)
if done_procs:
for p in done_procs:
del proc_dict[p]
time.sleep(2)
edit: if I proc.communicate()[0] in the else: where the sleep is located, some.exe is able to continue/finish, but as communicate waits for the process, it slows down the script way too much.
