[R-pkgs] New Matrix and lme4: Must reinstall lme4 if got new Matrix

Martin Maechler maechler at stat.math.ethz.ch
Wed Aug 10 17:28:35 CEST 2011

We have released to CRAN a new version of the (recommended)
package Matrix, and of package lme4 yesterday.

Anyone who gets the new version of Matrix *MUST* re-install lme4
-- if (s)he is using lme4 at all.  Technical details about that further below.

The fact that yesterday's version number of Matrix is 0.9995875-2, 
indicates that Matrix' version is indeed approaching 1.0 (*),
and I'd declare this version as  "release candidate" for 1.0.
As a recommended package, Matrix is part of every R distribution,
our aim is to release Matrix_1.0-0 (or higher) with the next
non-patch release of R, i.e., R-2.14.0 somewhere in October.

For this reason, we are asking R useRs, programmeRs and
provideRs, to ``hash at'' the package, trying to find problems /
bugs, badly lacking (or wrong) documentation, etc,
and report it to us (to 'Matrix-authors at ...' or possibly R-devel at ...),
so we can prepare a  shiny sparkling  Matrix_1.0-0 in time.
Thank you in advance for such a contribution to the Free
Software universe.

Martin Maechler and Doug Bates
(and Ben Bolker for lme4).

-- -- --

- Why must lme4 be re-installed as well ?  [Technical !]

 This is because lme4 has a 'LinkingTo: Matrix' in its
 DESCRIPTION and indeed, lme4's C code is using part of Matrix' C
 The change in Matrix: Part of the C-level interface in Matrix is
 now (again, after several years) using the standard CHOLMOD
 typedef of UFlong.  This means that the C API of Matrix should
 behave conformly with what other instances of CHOLMOD export.

(*) For some, it may be amusing to read
    which was the last announcement of Matrix on R-packages.

More information about the R-packages mailing list