On Thu, Aug 26, 1999 at 10:59:30AM -0500, Brandon W. Beasley wrote:
| What are ways to detect bad memory?
|
| Does a successful kernel compile usu. inidicate good memory? Will
| compiling a kernel usu. sniff upchuck on bad memory?
It's a good start, but many memory problems are very sporadic. So one
compilation isn't good enough. In fact, what I normally do is do
kernel compiles over and over and over for several days, perhaps as
long as week. If it doesn't fail, it's a pretty good bet that your
memory is good. It's not bulletproof, but nothing is.
I use the following script to compile over and over and over, aborting
if there are any errors :
#!/bin/sh
check () {
error=$?
if [ $error -ne 0 ]; then
echo "error $error - aborting at `date`, during compilation $COUNT"
exit 1
fi
}
COUNT=0
while : ; do
COUNT=`expr $COUNT + 1`
echo "=== Starting compilation $COUNT at `date` ..."
make clean
check
make depend
check
time make -j2 bzImage
check
echo "=== Done with compilation $COUNT at `date` ..."
sync
sleep 5
done
{ yes, my box is SMP. }
I guess if I really wanted to get slick I could compare kernels to the
last one and see if they match - but due to ELF they may not match
even though both compilations succeeded. Either way, I've not done it
and probably won't.
--
Doug McLaren, [EMAIL PROTECTED]
---------------------------------------------------------------------------
Send administrative requests to [EMAIL PROTECTED]
