Re: bizarre code from logs-to-c

[Justin Mason <jm...@jmason.org>]

-----BEGIN PGP SIGNED MESSAGE-----
Hash: SHA1


Daniel Quinlan writes:
> Does the last part of this patch make *any* sense?  It's Allen code from
> January 2003 (bug 1088), checked in without a comment probably because
> none of us understood it.  (The single line changes in the patch are
> fine.)
> 
> As far as I can tell, all it does is screw up scores when doing
> fp-fn-statistics because read_ranges now happens before readlogs at the
> beginning of logs-to-c.  So, some scores are *changed* before doing
> counts for fp-fn-statistics.  Before the move, it looks like a complete
> no-op, but it's been this way since mid-2004.
> 
> This change confuses the hell out of me.  I can't even figure out what
> the hell Allen was *trying* to do.

Agreed; it's entirely incorrect.  "logs-to-c" should not be modifying the
contents of the %scores hash at all in that function, since as you say, it
screws up fp-fn-statistics.  :(

that's not good news.... have our FP%/FN% rates been mis-measured
by this?

- --j.
-----BEGIN PGP SIGNATURE-----
Version: GnuPG v1.4.1 (GNU/Linux)
Comment: Exmh CVS

iD8DBQFD2R6aMJF5cimLx9ARAncaAKCizEP04+naeK1TSHmsxD86IFTIdgCffZGn
l4k1LUm4q4QQ2JHaRE7MFDQ=
=nUIz
-----END PGP SIGNATURE-----