Tape driver in my pc532 kernel seems to be misbehaving 
	( kernel is 94 04 29.sd0.GENERIC_Z	cksum=3581997180 348176)

I several times wrote tapes (varying the media) on my FreeBSD i486,
sometimes with dd ( and conv=sync), 
sometimes with tar (default 10k) & sometimes with -b 8k
each time the last bit of the file was garbled after reading on PC532 !
(I tried reading back to i486, & cmp, no problem all OK that end).

Possibly with a normal tar one might not notice, but I was moving
a tar.gz (size 22804131) of all NetBSD current sources,
so I guess gzip caught the corruption.

In desperation I finally used split on my main host (i486) & tar'd all the
xaa xab etc to tape, still the last file (xdi) had a different cksum,
so I kermited that across.  The main xaa xab were around 260K each,
& identical on each machine, the difference was in the final small 22K xdi file.

I examined the difference: 1024 bytes had been forced to 0, 62 bytes before the end of the file (ie last 62 bytes were OK)

As a temporary expedient, perhaps if one hangs a 50K file of pre gzip'd
dummy data at end of tape, it'll maybe fool the tape driver ?

PS I still haven't compiled a kernel of my own yet,
   but got fresh mirrored code Wed. June 15, so am hoping...
   (That code is what I was using tape drive for)

PPS Stuart has also noticed problems when not using 8K blocking.
	on his kernel (cksum reports 2431357043 348176 940429.0.s.G_Z )
	this kernel having been specially made by phil sometime back
	for stuart's impoversished hardware no 6250 no dallas rtc )

Julian H. Stacey	Munich, Germany		Tel. +49 89 268616  TZ=GMT+1
 <stacey@freefall.cdrom.com>		California, forwarded.
 <julian@pcsbst.pcs.com>		Munich, Fridays.
 <jhs@reseq.e-technik.tu-muenchen.de>	Cease Use.
 <stacey@guug.de>			Cease Use, Expires 30th June '94.

------------------------------------------------------------------------------