something strange with amd64 or fetch or ports or gcc

While trying to build math/octave, make(1) stopped with

lapack.tgz                                     99% of 4874 kB 2448 kBps
fetch: lapack.tgz appears to be truncated: 4991991/4991992 bytes
>> Couldn't fetch it - please try to retrieve this
>> port manually into /usr/ports/distfiles/lapack and try again.
*** Error code 1

Stop in /usr/ports/math/atlas/work/lapack.
*** Error code 1

Stop in /usr/ports/math/atlas.

I simply restarted make(1) and the port infrastructure computed and
validated the md5 checksum on lapack.tgz without me manually retrieving
the tar ball.  So all 4991992 bytes were successfully retrieved from
netlib.org.

The system is a SMP amd64 system, and FreeBSD (world and kernel) were
rebuild from today's sources.  I use the standard CFLAGS of "-O -pipe".

I haven't seen other reports of this behaviour.  This could
be amd64 specific, caused by the new gcc, a 64-bit issue with
fetch (I doubt this it, but...), or some problem in the 
ports system *.mk glue.

-- 
Steve