At 2:12 AM +0900 5/3/05, Izumi Tsutsui wrote:
>
>Does the problem also happen without -j or not?

As far as I know, but it might not be the exact failure.  After six or
seven unsuccessful builds in the last two weeks and several more since
January, these are starting run together.  Primarily I am trying to
determine how local to me is the problem.

>> The most recent error occured with -j 8 and a 2.99.10 kernel, the most
>> recent I could find that would reliably run.  Previously I had run all
>> three -j variations with 3.99.3, none successfully.  Maybe the logs will
>> have something more informative.  The message I posted was from the
>> terminal.
>
>Are you sure you have enough memory and swap on your host?
>(Aren't there any kernel messages in /var/log/messages?)

I will check when I do the cvs up later tonight or tomorrow.  I think it
best for both me and the box if it remains turned off for a while.

>It requires a bunch of memory to compile icache.c and semantics.c
>which are generated under gdb/sim/ppc.
>---
>% ls -l sim/ppc/obj.macppc/icache.c
>-rw-r--r--  1 root  wheel  3546470 May  3 00:48 sim/ppc/obj.macppc/icache.c
>% ls -l sim/ppc/obj.macppc/semantics.c
>-rw-r--r--  1 root  wheel  2187998 May  3 00:48 sim/ppc/obj.macppc/semantics.c

This is an interesting point.  I did take a 128M stick out, leaving it with
128M, some time at or around the last successful userland build.  I can not
say which occured first.  This would differentiate the error I am having
now from the commits that don't compile.  If this indeed turns out to be
the problem, some note could be inserted in build.sh to issue a warning if
physical memory is too small and likely to have problems.

tim