🚀 go-pugleaf

Compiling various public domain programs on our NeXTstation using the
2.1 operating system and gcc as supplied with the machine has revealed
the following problem. Although the programs invariably pass any test
routines supplied with the program and will perform correctly in many
situations, we are plagued with segmentation faults and occasional bus
errors. Some users have also complained that their programs also
produce segmentation errors.

I beleive the problem is with memory allocation. We do lengthy
calculations typically requiring days to complete. One particular
program has a flag which displays statistics during garbage
collection. With this flag, we see that just prior to the segmentation
fault the program does garbage collection almost continually. I do not
beleive these programs are able to get more memory when they ask for
it and therefore exit with a segmentaion fault. All of the programs in
question have been successfully compiled on other platforms without
adverse behaviour including some of the user's programs. Indeed, all
of these programs will function properly with "small" inputs but fail
when attempting the lengthy calculations.

Are we experiencing a known bug? I've looked through the various
sources of information regarding the NeXT, but haven't found any
reference to these problems. Any help would be appreciated. I am a
novice C programmer so I would appreciate less technical responses
although I would like to understand the technical aspects. There are
people here who are "real" C programmers however.


Jackie Ford
University of Rochester
ffor@troi.cc.rochester.edu (non-NeXT mail)
ffor@gauss.math.rochester.edu (NeXT mail accepted)

--
Jackie