[fprint] pam_fprint improvements

[Daniel Drake]

Hi Vasily,

Sorry for the delay in getting to this. Congratulations (and thanks) for 
being selected to work on fprint as your university project.

Out of interest, do you have to do any writeup as well as the actual 
coding? If so, will you be publishing it?

For my own project I have to write a big report and plan to publish it 
under a creative commons license in early May.

> Sorry, last patch introduced regression for devices that do not support 
> identification. New one fixes that.

Can you combine this patch into one? "git commit --amend" and "git 
rebase --interactive" are helpful tools here :)

The patch seems a bit messy in that it creates a gallery even for prints 
that do not support identification, and it also is a bit confusing 
code-flow wise. For example do_identify() calls fp_verify_finger() on 
devices that don't support identification.

Any chance you could give it a bit of a rethink? I'd envision some kind 
of dual codepath system depending on whether the device supports 
identification or not, with some helper functions to simplify the common 
parts (such as interpreting results).

FYI, in the long term I am thinking about folding identification and 
verification into just one function (identify), and allowing drivers to 
place an upper limit on the gallery size. Devices that can't do 
identification would set this limit to 1.

Thanks!
Daniel