Re: [NSE] Unused library cleanup

[Daniel Miller <bonsaiviking () gmail com>]

Patrick,

Thanks for this info. It's a lot to think about, but I think I'll take 
your advice and work on nse_remove_require. Just to be sure that I 
haven't messed anything up, I'll hold off on committing my patch until 
I've had the chance to play around with the existing tools and better 
understand what I'm doing.

Dan

On 07/09/2012 03:33 PM, Patrick Donnelly wrote:
> Hi Daniel,
>
> On Sun, Jul 8, 2012 at 9:27 AM, Daniel Miller <bonsaiviking () gmail com> wrote:
> > I wrote a small Perl script (attached) that checks for Lua libraries
> > required but not used (for one definition of used).
> I suggest you maybe look at
> /nmap-private-dev/misc-scripts/nse_remove_require and
> /nmap-private-dev/misc-scripts/nse_check_globals.
>
> The first script removes all occurrences of require so that
> nse_check_globals can add and alphabetize all requires.
> nse_remove_require is also incidentally useful for removing unneeded
> requires. Unfortunately nse_remove_require needs some work as it
> currently removes require lines that it should not, e.g. commented
> requires:
>
> -- @usage
> --   require "foo"
>
> You may want to improve upon nse_remove_require so that it is more
> foolproof. So far, I know it should not remove require statements like
> this:
>
> o Lines which are commented (space then "--").
> o Move silent_require statements to top of script.
> o Don't remove instances of require "strict"
> o Don't remove requires that are commented with "cycle" (smb.lua)
>
> I played around with vim regular expressions and have given up. I
> think something like perl or Lua should easily work but haven't gotten
> around to fixing it.
>
> > 2. I'm not sure if the "method"-style of function use is affected by
> > having or not having a local reference to a library. For example,
> > string.format("%d", 3) runs faster if local string = require "string".
> > But is the same true for ("%d"):format(3)?
> No. The string type has a global metatable shared by all strings which
> is set by the Lua string library [1]. This is why it's possible to
> index string methods from a string. Whether or not you have a local
> reference to the string library is irrelevant. To illustrate:
>
> $ bin/lua
> Lua 5.2.1  Copyright (C) 1994-2012 Lua.org, PUC-Rio
> > print(require "string", getmetatable("").__index)
> table: 0x205aca0        table: 0x205aca0
> > print(require "string".format, getmetatable("").__index.format, ("").format)
> function: 0x41dad0      function: 0x41dad0      function: 0x41dad0
>
> [1] http://www.lua.org/source/5.2/lstrlib.c.html#createmetatable
>
> > 3. packetdecoders.lua had requires inside functions, which I think is
> > not a great idea, since those functions should try to be as fast as
> > possible. I moved them to the top of the file instead.
> Be careful because there are some libraries (smb.lua is the only one I
> know of) that need to do this. It looks like your fix for
> packetdecoders is good though.


_______________________________________________
Sent through the nmap-dev mailing list
http://cgi.insecure.org/mailman/listinfo/nmap-dev
Archived at http://seclists.org/nmap-dev/