Re: Patch that adds a new test to find

[Eric Blake]

[it's customary to reply inline, rather than top-posting, on technical
lists]

On 04/05/2010 03:13 PM, address@hidden wrote:
> Thank you very much for your kind and accurate answer!  I have tried to
> modify my patch so as to solve the problems you pointed at.  I know that
> looking at the content of a file is not in the spirit of find, but anyway
> I believe it is a very useful exercise for me to go and hack such an
> excellent piece of software as findutils is... and moreover I would be
> glad to bring some little collaboration to the GNU project.
> 
> As to the documentation and the regression tests, I have not yet worked on
> them, but surely will in the next days.

One other thing that you need to be aware of - your patch is large
enough that we cannot incorporate any of it into findutils unless you
first assign copyright to the FSF.  If this is something you are
interested in pursuing, then we can give you details off-list on how to
proceed.  The process typically takes a couple of weeks.

>>
>> I hope you find the feedback above helpful; I'm certainly glad to find
>> a volunteer willing to work on findutils.   Having said this, I'm not
>> certain that making such a check this way is really better than using
>> "-exec cmp".   Considering after all that this is a substantial amount
>> of extra code.   However, someone else from the list may well have an
>> opinion different to my own.

I can see the potential for savings here; particularly if the comparison
file can fit into memory, because then you not only have fewer processes
due to -exec cmp, but also the potential for fewer disk accesses (each
separate cmp process would reread the file, while doing it in find could
just reuse the one in memory).  I don't know if you are taking the
shortcut of comparing file size, and skipping the comparison altogether
if the candidate file and the comparison file are different sizes, but
that is certainly an optimization worth considering (although for
non-regular files, such as in /proc, file size is not an accurate measure).

But it's still not something that I've ever found myself needing to do,
so I'm still on the fence on whether this makes a good addition.

-- 
Eric Blake   address@hidden    +1-801-349-2682
Libvirt virtualization library http://libvirt.org

signature.asc
Description: OpenPGP digital signature