ddl@wjh12.UUCP (lanciani) (04/18/84)
[] We have been trying to run plain vanilla unix v7m on a PDP-11/44 with two RK07 drives (no floating point). At first, everything appeared fine; but then we noticed that any program that created a pipe crashed the system. The crash occasionaly took the form of a panic: trap, but usually the CPU would just halt and return to console mode without saying anything. Careful probing with pstat seemed to reveal that the inode was created correctly by pipe() and that data writen was indeed buffered; reading caused the crash. In an attempt to isolate the problem, I added printf() debuging statements to the kernal pipe() code. Unfortunately, the problem vanished. Replacing the printf()'s with nulldev()'s worked as well. However, changing the length of the debug strings caused the system to crash in various ways (panics, halts, etc..). DEC has run memory diagnostics and has replaced every board in the CPU cage (one at a time) to no avail. It sounds like a memory problem, but we can't find it. Does anyone know of some strange bug that could account for this? Please mail me directly if you have any ideas. Thank You. -- From the TARDIS of Dan Lanciani (ddl@wjh12) ...!decvax!genrad!wjh12!ddl