os.fdopen(9, 'w') about argcomplete HOT 5 CLOSED

Ah, great point, thanks. Your alternative way sounds like the way to go.

from argcomplete.

OK, I'm confused. I admit that I don't know precisely how fd management works in bash. From reading http://www.tldp.org/LDP/abs/html/io-redirection.html, the only conclusion I have been able to make so far is that "9>&2" in https://github.com/kislyuk/argcomplete/blob/master/argcomplete/bash_completion.d/python-argcomplete.sh#L31 does cause bash to reserve fd 9, by connecting it to fd 2, which is already open.

Could you help me understand what I need to do to make sure fd 9 is reserved correctly?

from argcomplete.

The time when it came up is actually when not autocompleting. My advice
was off, sorry!

The problem was because importing argcomplete causes it to try and open
that descriptor whether autocomplete() is called or not.

I fixed it locally by having it use stderr whenever _ARGCOMPLETE is not
set.
On Dec 14, 2013 4:06 PM, "Andrey Kislyuk" [email protected] wrote:

OK, I'm confused. I'm a bash n00b so I don't know precisely how fd
management works in bash. From reading
http://www.tldp.org/LDP/abs/html/io-redirection.html, the only conclusion
I have been able to make so far is that "9>&2" in
https://github.com/kislyuk/argcomplete/blob/master/argcomplete/bash_completion.d/python-argcomplete.sh#L31
does cause bash to reserve fd 9, by connecting it to fd 2, which is
already open.

Could you help me understand what I need to do to make sure fd 9 is
reserved correctly?

—
Reply to this email directly or view it on GitHubhttps://github.com//issues/66#issuecomment-30588687
.

from argcomplete.

Aha! Thanks for the clarification. The bug is something different, but just as severe: it shouldn't open any fds when imported. I'll fix that asap.

from argcomplete.

Should be fixed in f4c3600 and released in v0.6.4. Please test.

from argcomplete.