Re: [buildd] building r-base
> > hobbes froze yet again building dwm - I'll leave it off for now until I
> > find time to check it out (might take two weeks or so).
>
> ahhz froze building mcmcpack_0.8-1-1. I tried to capture a dmesg, but it
> seems it did not get saved. However, I have one from the previous freeze,
> which looks similar. There were hundreds of esp0 messages at the end:
>
> IPv6 over IPv4 tunneling driver
> process `syslogd' is using obsolete setsockopt SO_BSDCOMPAT
> vt: argh, driver_data is NULL !
> NETDEV WATCHDOG: eth0: transmit timed out
> hda: status timeout: status=0xff { Busy }
> ide: failed opcode was: unknown
> hda: drive not ready for command
> ide0(?): unexpected interrupt, status=0xff, count=1
> esp0: Gross error sreg=7f
> esp0: No current cmd during gross error, resetting bus
> esp0: Resetting scsi bus
> esp0: Gross error sreg=7f
> esp0: No current cmd during gross error, resetting bus
> esp0: Resetting scsi bus
> esp0: Gross error sreg=7f
> [...]
>
> My ide controller is a Catweasel (Xsurf), thats where the whole system is
> installed. esp0 is a Blizzard2060, but the only partition on there that
> should be used is the swap partition. Is R trying to use obscene amounts of
> (swap) memory?
That might be the cause - OTOH I do not recall such anomalies from
previous hangs. And the OOM killer should trigger, anyway.
We need to have something to monitor VM usage reliably. Plus network
logging, perhaps. If you can move swap off SCSI (or just disable swap) we
can figure out if R somehow triggers SCSI bugs :-)
Neither SLIP nor Ethernet worked on hobbes last time. I'll reconfigure the
serial line for console/login now.
Another thought: might the EtherNAT act up on me??
Michael
Reply to: