Re: Unexpected soft update inconsistency

[walt <wa1ter@xxxxxxxxxxxxx>]

On Fri, 9 Jun 2006, Matthew Dillon wrote:

>
> :> SIZE=512 MTIME=[date] COUNT 21 SHOULD BE 22
> :> LINK COUNT INCREASING
> :> UNEXPECTED SOFT UPDATE INCONSISTENCY
> :>
> :> I tell it to correct the error and all seems fine until the
> :> next 'write error' while trying to build firefox.  The error
> :> always occurs just after the message 'leaving directory <foo>'
> :> so it's not entirely random the way an overheating CPU might
> :> behave.
> :>
> :> Any ideas how I can debug this?
[...]

>     It might be possible to trace out the cause of the bmake write error
>     by ktraceing the build.  Something like this:
>
>     ktrace -i -t cns bmake

Heh.  I tried your suggestion and I used up 2.2 Gigs of disk space
and then got a (legitimate) 'disk full' error.  ktrace.out had
grown to 2GB.

The bmake of firefox usually runs for only five minutes or so
before crapping out with 'write error' -- so my gut tells me
that using ktrace probably prevented some race condition from
causing the 'write error'.  But of course I'm only guessing.