Thanks armadillo 1.99.2 -> 1.99.3

[Dirk Eddelbuettel]

Dudes,

Just saw this bubbling up in the RSS feed of new packages.  

I keep a pretty close eye on this as my Rcpp (R and C++ integration package;
also in Debian) has a nice related package RcppArmadillo (not yet in Debian;
I should probably change that though it is complicated -- see below) for
templated use of Armadillo with R via Rcpp.  And I have a pretty decent
relationship with upstream because of this.  You may want to upgrade to
1.99.3:

   From: Conrad Sand <conradsand at gmail.com>
   To: Dirk Eddelbuettel <edd at debian.org>
   Cc: "Romain, Dirk and Doug" <RcppArmadillo-authors at r-enthusiasts.com>
   Subject: Re: CRAN submission RcppArmadillo
   Date: Fri, 27 May 2011 22:49:23 +1000
   
   Hi Dirk,
   
   Several users reported a few problems with 1.99.2.  The main issue was
   empty matrices being too readily interpreted and converted into empty
   vectors.  There was also an issue with .diag(), also involving empty
   matrices.  For such as "empty" feature, it's definitely a friggin'
   pain.
   
   I've released 1.99.3:
   http://sourceforge.net/projects/arma/files/armadillo-1.99.3.tar.gz
   
   As a sweetener, there's a speedup in dot() when dealing with submatrices.
   I've also disabled the support for u64, so we don't have to deal with
   the "long long" problem. It also caused problems on 32 bit machines.
   I'll reintroduce it once C++0x is official and GCC supports it
   properly (1 year from now?)
   
   It's a bit annoying to release it so quick after 1.99.2, but I'd
   rather not worry about it over the weekend.
   
   Cheers,
   Conrad

As for the 'why it is complicated': RcppArmadillo 'knows' it is used within R
so I don't have to run cmake -- we do not ship a shared library either
reducing this to pure template use which is extremely handy for us: no
depends on Windoze, OS X, other Linux distros, ... 

Dirk

-- 
Gauss once played himself in a zero-sum game and won $50.
                      -- #11 at http://www.gaussfacts.com