As of a recent revision (39064 in svn), the problems we've been able to
isolate in the CRAN packages seem either to be solved or not obviously
related to the methods changes. If anyone has a reproducible example
that seems to point to the changes, let me know.
One more change is planned, likely for tomorrow: We will introduce a
bit in the internal data structure that is turned on for S4 objects,
giving a quick and reliable test. Compared to the last round of
changes, this should have little direct impact, but it will allow more
reliable and efficient handling of S4 objects. It will, however, also
require objects to be created with the new code in order for the test to
work. So don't file away your saved data images just yet, if they
contain S4 objects they will need to be recreated.
John
