Greetings folks: Stepped away from a win/lin dual boot system and spent the better part of the past week resetting a windows only arrangement. Before the switch I was running 2.3.1 without any strange behaviours. When I reached the point of putting it back on the machine I would get this puzzling warning the first time I would try to select a cran mirror: > chooseCRANmirror() Warning message: DLL attempted to change FPU control word from 8001f to 9001f > When I pulled it off and reinstalled 2.2.1 I got no such warning. Thoughts or suggestions? Could the file have been damaged or corrupted in same way? The Md5sum checked out but I'm not sure how accurate an assessment that is of the file's integrity. Should I just download a new copy and try again with that? Not sure if it's important but I run a Win98se setup. -- Brian Lunergan Nepean, Ontario Canada ---
On Sun, 13 Aug 2006, Brian Lunergan wrote:> Greetings folks: > > Stepped away from a win/lin dual boot system and spent the better part of the > past week resetting a windows only arrangement. > > Before the switch I was running 2.3.1 without any strange behaviours. When I > reached the point of putting it back on the machine I would get this puzzling > warning the first time I would try to select a cran mirror: > > > chooseCRANmirror() > Warning message: > DLL attempted to change FPU control word from 8001f to 9001f > >This means that some other program (often a video driver) rudely changed the floating point control word, but that R caught the change and stopped it. From R's point of view the problem was prevented. Now, it is possible that the other program thinks the floating point control word is still set the way it wants, in which case it might have problems with numerical accuracy. R should be ok, though.> When I pulled it off and reinstalled 2.2.1 I got no such warning. > > Thoughts or suggestions? Could the file have been damaged or corrupted in same > way? The Md5sum checked out but I'm not sure how accurate an assessment thatis > of the file's integrity.If the MD5 hashes agree then the files are the same as when the installer was built. There is negligible probability that a change in the file will leave the MD5 hash the same. -thomas Thomas Lumley Assoc. Professor, Biostatistics tlumley at u.washington.edu University of Washington, Seattle
On Sun, 13 Aug 2006, Brian Lunergan wrote:> Greetings folks: > > Stepped away from a win/lin dual boot system and spent the better part > of the past week resetting a windows only arrangement. > > Before the switch I was running 2.3.1 without any strange behaviours. > When I reached the point of putting it back on the machine I would get > this puzzling warning the first time I would try to select a cran > mirror: > > > chooseCRANmirror() > Warning message: > DLL attempted to change FPU control word from 8001f to 9001f > > > > When I pulled it off and reinstalled 2.2.1 I got no such warning. > > Thoughts or suggestions?This is a warning about a problem with your OS, not with R.> Could the file have been damaged or corrupted in same way? The Md5sum > checked out but I'm not sure how accurate an assessment that is of the > file's integrity. Should I just download a newDid you run md5check?: that is very thorough. See the rw-FAQ Q2.3.> copy and try again with that? Not sure if it's important but I run a > Win98se setup.This is covered by rw-FAQ Q2.21 (although that will not be obvious to you, but please read it now). It indicates a problem with some *non-R* DLL on your machine, and not a problem with R itself. R 2.3.1 looks up the latest list of mirrors online (and is much more aggressively protected against rogue DLLs), and 2.2.1 did not, so my guess is that it is your winsock or other internet access DLLs. R has corrected the problem for you. Your OS have been end-of-lined by Microsoft long ago, but I do suggest you ensure it is as fully updated as possible (including Internet Explorer), assuming that using a current OS is out of the question. (R support for that OS is likely to be withdrawn in the near future: it was somewhat unexpected that 2.3.x still worked on Win98 since it uses features that were said in some accounts not to be present.) -- Brian D. Ripley, ripley at stats.ox.ac.uk Professor of Applied Statistics, http://www.stats.ox.ac.uk/~ripley/ University of Oxford, Tel: +44 1865 272861 (self) 1 South Parks Road, +44 1865 272866 (PA) Oxford OX1 3TG, UK Fax: +44 1865 272595