Re: toolchain/44096: gcc crashes while compiling gsromfs1.o

[Ignatios Souvatzis <is%netbsd.org@localhost>]

On Tue, Nov 16, 2010 at 06:12:56AM +0000, David Holland wrote:
> On Mon, Nov 15, 2010 at 11:30:00PM +0100, Ignatios Souvatzis wrote:
>  > > > gcc: Internal error: Killed (program cc1)
>  > >  
>  > >  That means signal 9 - are you out of swap?
>  > 
>  > 9? shouldn't OOS create SEGV instead?
> 
> Depends how it strikes maybe, but usually not:
> 
>       printf("UVM: pid %d (%s), uid %d killed: "
>           "out of swap\n", l->l_proc->p_pid, l->l_proc->p_comm,
>           l->l_cred ? kauth_cred_geteuid(l->l_cred) : -1);
>       ksi.ksi_signo = SIGKILL;
> 
> That's in sys/arch/arm/arm32/fault.c; other arches are similar.
> 
> Check directly with dmesg, because IME when you run out of swap one of
> the first processes to get this treatment is almost always syslogd. :-/

No, syslogd is still running, and didn't get that kill AFAICT:

UVM: pid 438 (beam), uid 1501 killed: out of swap
UVM: pid 6020 (fetchnews), uid 1003 killed: out of swap
UVM: pid 1049 (fetchnews), uid 1003 killed: out of swap
UVM: pid 17862 (fetchnews), uid 1003 killed: out of swap
UVM: pid 6476 (fetchnews), uid 1003 killed: out of swap
UVM: pid 23404 (fetchnews), uid 1003 killed: out of swap
UVM: pid 10721 (cc1), uid 0 killed: out of swap
UVM: pid 12065 (fetchnews), uid 1003 killed: out of swap
UVM: pid 18124 (cc1), uid 0 killed: out of swap

I wish the error message on the tty was more clear, but I understand
this is not really possible in this situation. 

        -is