> From: Andrew Gillham <gillhaa@ghost.whirlpool.com>
> Hello,
>
> On my Multia with 1.4.2_ALPHA I am unable to complete a make build, as
> eventually I'll get:
> cc: Internal compiler error: program cc1 got fatal signal 11
>
> This is with a GENERIC kernel.  If I boot a 1.4T GENERIC kernel and
> use the 1.4.2_ALPHA userland, the make build completes without any
> problems.  Has an internal kernel limit changed that cc1 is bumping
> up against?  Or is this more likely bad memory or something?
>
> I suspect it is related to the SRM console though, as I recently updated
> my SRM to the V4.5-819 mentioned WRT running VMS.  NetBSD failed to boot
> with this version (cache errors in 'ncr', and DMA(?) errors with 'de')
> installed.  I then realized I was running an interim release (V4.5-780?)
> that I didn't have anymore. :(  Stupid me.  Anyway, so now I am running
> the "BLT5 V3.8-3" that is available from the DEC ftp site.  It has a number
> of problems compared to the newer versions, but boots ok from dka0.
>
> So, I guess I have two questions:
> 	Is the 1.4.2_ALPHA .vs. 1.4T issue likely to be the kernel, the SRM,
> 	or my hardware?
>
> 	Has anyone had any success using SRM V4.5-819 with NetBSD?

This is a usually an easy question to answer, however your dependency is
a new twist.

Almost always, a transient build failure like that is a HW problem if
continuing the build "works", in the sense that the same file/module will
build and then another sig 11 will happen (unpredictably) down the road.

However, there is always some finite possibility of a SW or FW issue.

Is the behavior sufficiently reproducible to verify your stated dependency?
That is, is the MTBF << time_spent(testing 1.4T)??

	ross