dyer@spdcc.COM (Steve Dyer) (11/28/90)
One of our developers has been the victim of his PS/2 crashing several hours after initiating a VERY long compile job. The details of the crash aren't relevant to this question. He noted to me that his make.log, which should ordinarily be quite full, is almost always 0 length or is deleted by fsck upon reboot. This is makes the crashes even more annoying than they should be, since he loses any ability to easily see how far he's gotten. What I can't figure out is why a file which has been open for writing for more than two hours, and which is steadily being appended to, should, after a panic, reappear as length 0. Certainly syncs have been occurring at some regular interval, and anyway, the amount of file activity would eventually cause MOST of the bdwrites to end up on the disk, no? I then realized that AIX PS/2 doesn't seem to have either the V.3 "bdflush" kernel sync daemon or its user-mode predecessor, /etc/update. Where is the equivalent feature hidden? Not that I think that's the solution to the problem. Is this somehow tied up with the weird Locus "commit" handling? I should note we aren't running TCF here. -- Steve Dyer dyer@ursa-major.spdcc.com aka {ima,harvard,rayssd,linus,m2c}!spdcc!dyer dyer@arktouros.mit.edu, dyer@hstbme.mit.edu