For the record:
I had an inquiry about a problem with the function lm.fit.recursive
in my quantreg package on a 64bit gentoo linux system. After
some debugging the problem was traced to the fact that on this
system g77 took the view that the statement:
data one/1.0d0/
meant that the symbol one should take the value zero. Replacing such
data
statements with their parameter statement equivalent, resolved
the problem. This obnoxious behavior seems not to be news to the gcc
folks,
see http://gcc.gnu.org/ml/gcc-bugs/2002-08/msg00195.html
but they don't show any inclination to fix it either. I am in the
process of updating my packages and thought that others might
appreciate a warning -- or,more likely, enjoy the schadenfreude of
enlightened
C-aphones.
url: www.econ.uiuc.edu/~roger Roger Koenker
email rkoenker@uiuc.edu Department of Economics
vox: 217-333-4558 University of Illinois
fax: 217-244-6678 Champaign, IL 61820
