davea@vlsi.ll.mit.edu (Dave Allen) (02/23/90)
Last week, I wrote that I had released a new version of my continental drift simulator. I posted it to comp.sources.misc, and many people wrote to me with questions. Here are some common answers: If you don't know what comp.sources.misc is, it's a usenet news group just like rec.games.frp. Use your news reader to subscribe to this group, or ask a local expert. If you were unable to locate the programs on comp.sources.misc the first couple of days after I posted the announcement, please look again. I posted the announcement and the source at the same time, but programs take a few days to clear the c.s.m moderator. The archive numbers are v10i077 and v10i078. You can get the source code by ftp. Several sites archive comp.sources.misc. For example, I looked at uunet.uu.net, and found the program in the directory comp.sources.misc/volume10/tec, with the filenames part01.Z and part02.Z. Ask a local expert for help in finding a local archive site or using ftp. Several people have written to me that the IBM PC version fails with a message "PANIC: infinite loop in pfree". This message implies that the linked list of plates is getting munged. Most people see the problem within a small number of iterations, like two or three, and it shows up every time they run the program. I can run the Sun version 100 times for 100 iterations each and never see the message. I would like some help in tracking down this machine-dependent bug. Here is a clue provided by one reader: From: Jeff Butterworth <butterwo@cs.unc.edu> > Eventually I tried fiddling with the compiler options again, and found that > setting the default character type to "unsigned" did the trick. The program > runs now, but it seems to be creating initial continents that are quite > squarish, unlike the ones I saw when leaving default characters as "signed". > Maybe some random number generator is now being messed up because it assumes > that characters are signed. Since I have no access to a PC here, I can't look for the bug. However, I would like to help somebody track it down and squish it. Please write to me and I will answer any questions you have about the code. Please e-mail me directly rather than posting, since I don't read all the groups where this message is going. Dave Allen: internet davea@vlsi.ll.mit.edu