To:	djgpp AT sun DOT soe DOT clarkson DOT edu
Subject:	Problems
Date:	Mon, 11 May 92 17:05:49 +0100
From:	"S. Linton" <sl25 AT cus DOT cam DOT ac DOT uk>
Status:	O

There seem to be TWO problems which are causing much grief and confusion.
I thought it might be worth a short posting to clarify and distinguish them
(at least as I encounter them).

1) Segmentation error or similar (error 14)
This seems to occurr only with 1.06, and specifically with the 1.06
extender. I have encountered it when compiling certain files and when
running various quite large applications. It may well be related to swapping,
but I can't prove that. It is repeatable, in that running the same program
on the same data will produce the same error, but not (easily) isolatable
in that almost any change to the input or program will remove or dramatically
change the error. I have avoided this by moving back to the 1.05 extender.
It may have been (in my case at least) related to swapping on a badly 
fragmented partition.

2) Hard disk failure - typically with abort retry fail message.  This
seems to occurr with 1.06 and 1.05 and to be related to intensive disk
activity. This is completely random (apparently).  Re-running the
program will usually avoid the error. There are some reports of this
occurring with other programs not using go32 at all.

My system is pretty much minimal for using DJGPP, which may account for
my getting some of these more frequently - 
	386SX 16MHz, 4MB RAM (was 2 for a while).

		Steve Linton

PS the decrease in compiler speed at gcc 2.1 is a big disappointment.
Might it be possible to either 

a) have the system() call notice if the running program is small and put
it in a chunk of memory reserved from DOS in the usual way if it is.

b) have a new version of the system call that runs a.out files only,
using the same copy of the extender and avoiding swapping out unless needed.

Either would make it possible to have a gcc program that was not swapped out
three times during a compilation.

- Raw text -