patrick keshishian wrote:
quoted text
> I recently attempted an installation from a usb memory stick, where
> installation of xfont47.tgz failed consistently due to "crc error"[0].
> The stick was prepared following steps from OBSD FAQ[1].
> 
> I recopied xfont47.tgz file to the stick, and confirmed (or so I
> thought) that the source and new copy of this file were identical
> using cmp(1).

Try renaming the failed file but leave it on the media, then make a 
fresh copy to the stick. That way you ensure that the (known) bad spots 
are left unused.

quoted text
> Restarting the install bombed on the same file. So I stuck the stick
> in my source computer and mounted it and ran md5(1) on it. Sure enough
> the checksum is different from the source. I did a umount(8) and a
> mount(8), then ran md5(1) again, and this time yet a new value was
> reported. While mounted, running md5(1) again reports the same value
> -- i assume this is because the file is cached at this point --
> explaining why cmp(1) initially failed to raise a flag. Every
> umount(8), mount(8) and md5(1) gives a new result:

And, lesson learned here I hope, always unmount/mount a media to improve 
the odds of catching an error when verifying against the original. It is 
as you suspect very likely that all or some of your copy was cached the 
second time you accessed it.

quoted text
> $ md5 /mnt/umass/usr/build/rel/xfont47.tgz
> MD5 (/mnt/umass/usr/build/rel/xfont47.tgz) = 75e6065782783455bfcf912ec8822ec3
> $ sudo umount /mnt/umass/
> $ sudo mount /dev/sd0a /mnt/umass/
> $ md5 /mnt/umass/usr/build/rel/xfont47.tgz
> MD5 (/mnt/umass/usr/build/rel/xfont47.tgz) = 3912b5506b63c105a945fe7535d6e70e
> $ md5 /mnt/umass/usr/build/rel/xfont47.tgz
> MD5 (/mnt/umass/usr/build/rel/xfont47.tgz) = 3912b5506b63c105a945fe7535d6e70e
> $ sudo umount /mnt/umass/
> $ sudo mount /dev/sd0a /mnt/umass/
> $ md5 /mnt/umass/usr/build/rel/xfont47.tgz
> MD5 (/mnt/umass/usr/build/rel/xfont47.tgz) = 315233c616a824057b2558d678aed2c4
> 
> This is a bit of an odd behavior to me. Does anyone have an
> explanation for this?

Yes, and you are clueing yourself in on it - the stick is failing. :-)

I'd recommend tossing the stick out the window and getting yourself a 
new one. Now, if that one fails too in a similar way, then you have a 
problem...

quoted text
> Incidentally, should the install script not catch the error condition
> and warn the user of the incomplete install at the next prompt?

Provided the install script saves state, which I don't know if it does. 
Someone else should probably comment on that. :-)


Regards,
/Benny

-- 
internetlabbet.se     / work:   +46 8 551 124 80      / "Words must
Benny Lvfgren        /  mobile: +46 70 718 11 90     /   be weighed,
                     /   fax:    +46 8 551 124 89    /    not counted."
                    /    email:  benny -at- internetlabbet.se