Date: Sun, 10 May 1992 11:21:05 +0200
From: Ove Ewerlid <ewerlid AT pinus DOT slu DOT se>
To: djgpp AT sun DOT soe DOT clarkson DOT edu, joachim AT lysator DOT liu DOT se
Subject: HD problems
References: <199205092205 DOT AA09788 AT robert DOT lysator DOT liu DOT se>
Mailer: VM 5.32 (beta) for GNU Emacs 18.57.6
Status: O



> Now, the problem was *not* djgpp and *not* Hyperdisk as it turned out. It was
> simple the number of wait states my computer used while accessing DRAM. When

Increasing the number of wait states can cause non valid assumptions 
about timings in interrupt drivers etc to become valid.
The problem may still be the way the the interrupt drivers are implemented.
(Maybe at the BIOS level ...)

I've ran into the problem with hangups during heavy swapping.
The problem could be fixed by adding more memory so it was not correlated
to the software. (Such as - "-O in gcc is the problem" - it merely casues gcc
use more memory and swap more.)

Anyway, djgpp is not the only software package that is plauged by hangups
during heavy disk use. (On the computers I have, 486 AT 25, 4Mb RAM, 110 Mb disk.)

A couple of weeks ago I tried to compact my HD with Nortons Speedisk utility.
I had to run the program 3 times before it managed to compact the entire
harddisk without hanging after some arbitrary amount of time into the compactation.
(Note: I have backups of the HD but I didn't need them since Speedisk merely
       hanged and didn't crash the HD.)