[Bug-cpio] cpio v2.5 checksum error

[John Sutton]

Hi there

This is a bug report only, no fix is suggested, sorry about that ;-(

I am using v2.5 (plus the patch suggested by Holger Fleischmann on 2004/03/01 
which is in any case not relevant here because I am not using copy-pass 
mode), in this pipeline:

        find . -depth -mount -print0 | \
            cpio -o --null --block-size=32 -H crc | \
            $RSH "cd $MNTDIR && cpio -idum --block-size=2048" || \
            errorexit "$RSH \"cd $MNTDIR && cpio\""

to backup live filesystems onto a remote system.  It is (I assume!?) the 
second cpio in the pipeline - the "cpio -i" - which is reporting these errors:

cpio: dcc/grey_db: checksum error (0x2a847669, should be 0x2a845346)
cpio: dcc/grey_db.hash: checksum error (0xacca97f, should be 0xacca978)

The first thing to note is that 2.5 is a *massive* improvement over 2.4.2 in 
this context.  In fact, 2.4.2 is simply useless when files are growing! OTOH, 
using 2.5, instead of getting widespread corruption, I now just get a number 
of warnings of this form:

cpio: File <filename> grew, <N> new bytes not copied

(as expected for a live filesystem) but otherwise everything proceeds fine 
_except_ for the checksum errors noted above.

It appears that there is something unique about the way in which the 2 files 
grey_db & grey_db.hash are changing which exposes this bug in v2.5.  I do not 
see these checksum errors for any other files, and it occurs only 
intermittently for these 2 files.

I am copying this to the author of the DCC who might be able to shed some 
light on this but it seems pretty clear that this must be a bug in cpio ;-(  
Still, overall cpio is now a usable utility in this demanding context and I 
am very grateful for that!

Best regards

***************************************************
John Sutton
SCL Internet
URL http://www.scl.co.uk/
Tel. +44 (0) 1239 711 888
***************************************************