Subject: Re: pkg/15501: rsync development branch causes corruption with alpha server
To: Manuel Bouyer <bouyer@antioche.lip6.fr>
From: Greg A. Woods <woods@weird.com>
List: netbsd-bugs
Date: 02/28/2002 13:22:14
Seems there was still some serious corruption of at least one more file
on rsync.netbsd.org yesterday:

START:2002/02/27-18:44:14: cvs update /work/NetBSD/src
cvs [update aborted]: unable to parse /cvs/master/m-NetBSD/main/syssrc/sys/arch/sparc/sparc/autoconf.c,v; `author' not in the expected place
DONE:2002/02/27-19:01:05: updating /work/NetBSD/src
 

However it seems to have been corrected with last night's rsync (from
rsync.netbsd.org, again with 2.5.2 and without '-c' or '-W'):

$ ls -l -c /cvs/master/m-NetBSD/main/syssrc/sys/arch/sparc/sparc/autoconf.c,v
-r--r--r--  1 syssup  supadmin  227500 Feb 28 04:04 /cvs/master/m-NetBSD/main/syssrc/sys/arch/sparc/sparc/autoconf.c,v
$ cd sys/arch/sparc/sparc
$ cvs update -p autoconf.c > /dev/null                            
===================================================================
Checking out autoconf.c
RCS:  /cvs/master/m-NetBSD/main/syssrc/sys/arch/sparc/sparc/autoconf.c,v
VERS: 1.162
***************
$ 


Unfortunately RCS files have no real integrity checking -- there's not
even a simple 16-bit CRC kept on the entire contents (as SCCS does), let
alone also on the contents of each "text" section, so a complete
validation of a copy of the repository can only be made by comparing it
directly with the original in some way (and by assuming the original is
correct), which probably means by making another copy of the original
with a mechanism that has higher inherent integrity capabilities
(eg. maybe by tape or CDROM or such if the machines are in different
locations).



> Try: rsync -v -c -W rsync://rsync.fr.netbsd.org/NetBSD-CVS/pkgsrc/mk/bsd.pkg.defaults.mk,v /tmp/foo,v

Yes, that works, thanks!  rsync.fr seems to have good copies of all the
files I knew were previously damaged on the other servers.

I guess it doesn't matter what the collection is called just so long as
the instructions on the web page are correct, though for a true mirror
server one would expect even the collections to have the same name (and
of course the pathnames within the collection to be identical).

-- 
								Greg A. Woods

+1 416 218-0098;  <gwoods@acm.org>;  <g.a.woods@ieee.org>;  <woods@robohack.ca>
Planix, Inc. <woods@planix.com>; VE3TCP; Secrets of the Weird <woods@weird.com>