Subject: Re: compiler failure on sun4m
To: Chris Amthor <>
From: Joel Reicher <>
List: port-sparc
Date: 12/30/2002 07:37:38
> As I figured out while studying the archives, there are reports of
> "sudden compiler deaths" on sun4m machines running NetBSD 1.6. As
> reported further on, only the SS1, SS1+, and the IPC are affected.
> I experience somthing quite similiar, but I'm not shure it's related
> to this. While trying to install skipstone from the package
> collection, the GCC (2.95.3) stops whith an "internal compiler error"
> when processing mozilla. Unlike in the problems mentionend in the
> archives, the GCC doesn't receive any SIGSEGV or something, but halts
> for an obvious intrinsic reason. It then wants me to send a bug report
> to the GNU Project.

The random copmiler death while building the kernel (which I think is
the report in the archives you're referring to) doesn't always
mention a SIGSEGV for me when it dies on an ELC. But if I check for
a core dump, it's there - cc1.core.

What's this "obvious intrinsic reason"?

> Is this related to the other Problem? Did anyone of you experience
> something similiar, and eventually figured out a solution? Or is the
> GCC in 1.6 really broken?

The essential (and most annoying) aspect of the other Problem is that
it's *random*. As such it doesn't seem to be a compiler bug, but an
OS bug. It seems to emerge on large files, but repeated attempts can
still compile them correctly. Is your problem random or 100%
reproducible, i.e. deterministic?


	- Joel