File Corruption

[Paul Slootman]

On Mon 12 Sep 2005, Stefan Nehlsen wrote:
> > 
> > I'm going to have a look at that now :-/
> 
> FIRST: I do not know if the corruption where really caused by rsync!
> 
> I had made a copy of the corrupt tree and use it know to find out
> what kind of corruption ocurred.
> 
> I knew that there where problems installing packets with a i686
> installation of SuSE 9.2. So I searched the suspicous content of the
> first SuSE 9.2 DVD for *.i?86.rpm files for what rpm -K reported a
> corruption.
> 
> I got a list of 101 broken files.
> 
> I compared them with their corresponding non broken versions by using
> "diff -u" on textfiles created with "od -x".
> 
> A typical result (samba-3.0.7-5.i586.rpm) :
> --- /tmp/ok.txt 2005-09-12 16:26:03.749124638 +0200
> +++ /tmp/broken.txt     2005-09-12 16:26:04.112057459 +0200
> @@ -51805,7 +51805,7 @@
>  3126000 7c04 cfd3 0392 dbaa 2b3d 2412 be45 a3ac
>  3126020 5825 a6de 9703 c8e7 ea18 17fe bb8b b41f
>  3126040 1f3e 912e 19b7 f843 e64a 8d39 9509 031a
> -3126060 093f b647 71af 8d62 1159 fbd0 3e30 e36b
> +3126060 093f b647 71af 9d62 1159 fbd0 3e30 e36b

Hmm...

A colleague recently bought an AMD64 system. The system looked OK, but
compiling the kernel gave random segmentation violations, and now and
then the system hung. It turned out the memory was bad (even though a
24-hour run of memtest86 showed no problems).

Such single bit errors sound like random memory corruptions.
Try copying such files many times with cp and running md5sums of the
copies.


Paul Slootman