> This is always with a sysinst and tarballs built from -current. Every few
> times I get a good 'make build' of -current, I build a distribution and
> try installing it onto a second HD, to ensure that I could use it as a
> snapshot to restore my -current machine after a disaster.
> 
> If I perform the exact same "installation test" with 1.3.2, I get no such
> messages about "stray interrupt 7".
> 
> The only difference other than 1.3.2 bootfloppy+tarballs versus -current
> bootfloppy+tarballs, is that the installation media for 1.3.2 is ATAPI
> CD-ROM, and the media for -current is ATAPI ZIP 100.

There is a comment in /usr/src/sys/arch/i386/isa/isa_machdep.c
specifically about this subject: :-)

         * Stray interrupts on irq 7 occur when an interrupt line is raised
         * and then lowered before the CPU acknowledges it.  This generally
         * means either the device is screwed or something is cli'ing too
         * long and it's timing out.

The comment says when it happens, but not why.  It's always
worried me somewhat because I've seen quite a few IDE
disc drives exhibit this problem, and certainly from a long
time ago.  Anyhow, it could be that.

My question is, is it really a problem ? :-)  Can't easily
tell because the counter is static :-P and I've never
had an opportunity to un-static it on one of those machines
to see if the number of strays is anything to worry about.

--