[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]
Date: Wed, 23 May 2012 02:06:17 +0200
From: Frank Dittrich <frank_dittrich@...mail.com>
To: john-dev@...ts.openwall.com
Subject: Re: Fwd: bash auto-completion for john
On 05/23/2012 01:16 AM, Solar Designer wrote:
> I think that you should not actually invoke =LIST to determine possible
> completions because they're almost always single-digit anyway and
> because the list might not appear instantly.
>
> $ time ./john -pla=list &> /dev/null
>
> real 0m1.511s
> user 0m0.124s
> sys 0m0.964s
Thanks for pointing that out. I didn't know this would take so long.
I occasionally measured run time for the completion on my atom netbook
(by renaming _john to __john and adding a new _john with just one
statement: time __john), and I always detected run times of less than
0.2 seconds, even for the most complex logic.
That's why I guessed it is not worth the effort trying to compute some
intermediate results just once and buffering them in variables __john_*.
The bash completion script for git does this. But if I do it, I'll
probably providing wrong completions if a user switches between a
locally installed jumbo version of john and a system-wide official john
version and tries using bash completion for both in the same session.
Not sure if this ever happens and should be supported.
I guess (but didn't try it) if I try completion for different git
versions in the same session, I also might getting some invalid completions.
I removed the logic to try finding completions using --platform=LIST.
Instead, I just list
--platform=N --platform=LIST
as possible completions for --platform= (only if --platform= is a valid
option according to the usage output) and
--device=N --device=LIST
as possible completions for for --device=
I should suppress this for CUDA builds, as CUDA doesn't support
--device=LIST. Currently, I don't check whether it is a CUDA build,
since I don't know what --list=build-info shows in this case.
Instead, I use a workaround (check whether --platform= is a valid option).
Frank
View attachment "john.bash_completion" of type "text/plain" (18291 bytes)
Powered by blists - more mailing lists
Confused about mailing lists and their use?
Read about mailing lists on Wikipedia
and check out these
guidelines on proper formatting of your messages.
