riddle@woton.UUCP (Prentiss Riddle ) (07/28/87)
andys@genesis.UUCP (a.b.sherman) writes: || ...The problem alluded to in several of the postings was not a lack of || swap space but a bug in the kernel that caused the system to panic when || running uucico *WHILE RELATIVELY IDLE!!!!* This has been fixed, and the || fixdisk is available from the hotline. The symptom is unexplained crashes || (with uucico or uusched active) in the middle of the night. 4:00 AM was a || favorite time, leading it to be called "The 4:00 AM crash bug". Hmm. That doesn't match our experience. We suffered from the symptoms described (although unfortunately not just at 4:00 a.m.), but no readily available software fix solved the problem. Instead, as I mentioned in a posting last week, the crash happens for us only on machines with a revision J, K or L expansion board. (We suspect, but have been unable to prove, that machines with over 2 Meg of RAM are immune from the problem even if they have the bad boards.) We were told a couple of days ago by AT&T that they have halted production of these boards and will ship us some earlier revision boards shortly. In response to my earlier posting on the subject, one person wrote to me saying that he had found a software solution: || From: im4u!ut-sally!ihnp4!ihjhh!jhh || To: im4u!woton!riddle || Subject: #$*^%$%*&^)(*& Combo Cards || || I too have a combo card, revision J or greater, which has been causing me || much grief since I upgraded to 3.5. Heck, I even work for AT&T and || couldn't get a solution through official channels. The Hotline told me to || try to, "Find an old rev, these new ones have a problem that we can't || fix." Jonathan Clark (ihnp4!mtune!jhc) was nice enough to help me out, || sending me several new kernels between January and May or June. Finally, || I have a kernel that works. It came with a new cmb.o and wind.o. This is || not the fixed kernel on FIXDISK 2.0, but Jonathan said that it would || eventually become a FIXDISK. Perhaps you could have your repair personnel || call him if they haven't gotten the latest kernel into a fixdisk. I || understand their reluctance to change the board - they don't have any old || boards that work that they can swap for new boards. It took a long time || to find and fix the problem because all of the 7300 developers have very || early combo cards, and couldn't reproduce the problem. || || By they way, the hotline told me that the easiest way to reproduce the || problem is to hook a terminal to the combo card, and hit break several || times on the terminal. Eventually, the system will lock up. || || John Haller --- Prentiss Riddle ("Aprendiz de todo, maestro de nada.") --- Opinions expressed are not necessarily those of Shriners Burns Institute. --- riddle@woton.UUCP {ihnp4,harvard,seismo}!ut-sally!im4u!woton!riddle
dennis@quincy.UUCP (Dennis W. Koncelik) (07/30/87)
| In article <917@woton.UUCP|, riddle@woton.UUCP (Prentiss Riddle ) writes: | | Instead, as I mentioned in a posting last week, the crash happens for us only | on machines with a revision J, K or L expansion board. (We suspect, but have | | In response to my earlier posting on the subject, one person wrote to me | saying that he had found a software solution: | | || From: im4u!ut-sally!ihnp4!ihjhh!jhh | || To: im4u!woton!riddle | || Subject: #$*^%$%*&^)(*& Combo Cards | || | || I too have a combo card, revision J or greater, which has been causing me | || much grief since I upgraded to 3.5. Heck, I even work for AT&T and | || couldn't get a solution through official channels. The Hotline told me to | || try to, "Find an old rev, these new ones have a problem that we can't | | --- Prentiss Riddle ("Aprendiz de todo, maestro de nada.") | --- Opinions expressed are not necessarily those of Shriners Burns Institute. | --- riddle@woton.UUCP {ihnp4,harvard,seismo}!ut-sally!im4u!woton!riddle I'm not sure if this would be the exact solution to your problem(s) but it might be useful for you to obtain a copy of a Maintenance Release of the Curses Package anyhow. If you are experiencing a problem in an application package that uses curses you may have a problem with an older Release of the Programmer's Package under Version 3.5. Seems that AT&T accidently turned on a parity bit on the port when curses was first called with an initscr(). The problem occured to us when we started to go over to 3.5 and we noticed it only on certain combo cards (revision J was one I remember). After working with AT&T they found a fault in their software and the following fix was issued (available only by request from the AT&T Hotline): Curses Terminfo Programmers Package Maintenance Release 3.0.1L Ask for it by name. -- Dennis W. Koncelik ...seismo!uunet!swlabs!godfre!quincy!dennis -or- American LP Systems, Inc. ...cmcl2!phri!gor!helm!quincy!dennis -or- 1777-18 Veterans Memorial Hwy. ...ihnp4!chinet!icus!quincy!dennis Islandia, New York 11722 VOICE: +1 516-582-5525