On Thu, Nov 02, 2006 at 12:41:36PM +0100, Bart Martens wrote: > Hi Peter, > > http://bugs.debian.org/cgi-bin/bugreport.cgi?bug=296387
That's interesting: someone actually did have problems with par2 because of bad RAM. I never reported it upstream because I decided it was probably just bad RAM on the machine that generated the par set. > Do you still have this bug? Is it still reproducible? Can you help me to > reproduce it here? I only ever saw this with one data set, and I've been using par2 regularly since then. I think it's likely that the par set was generated on a computer with bad RAM, or something. (It was downloaded from usenet). I think par2 would detect errors in transmission, but I could easily imagine the redundancy data getting corrupted in memory, and then getting checksummed and written to the par files. par2 does a lot of XORs (I think) over large amounts of RAM, so it's exactly the sort of program that would be sensitive to bad RAM. I haven't read up on the math it uses, but if the math just determines which blocks to XOR, it won't detect an error in the XORed data. It is probably possible to implement a verify that actually does a repair, to detect if there is a bad data block somewhere. The trick would be to do it efficiently, maybe by doing a repair that required all the available par blocks and not trying to find errors that would cancel each other out in that case. Options for dealing with possibly-corrupt par blocks could also be considered (i.e. exclude block #15 from being used for repairs), or keep trying to repair excluding different sets of blocks until you get the right checksums on the output! I'll email again if I still have the data set, but I think I deleted it a while after I found a good copy of the file I was trying to get. Thanks for following up on this, BTW. -- #define X(x,y) x##y Peter Cordes ; e-mail: X([EMAIL PROTECTED] , des.ca) "The gods confound the man who first found out how to distinguish the hours! Confound him, too, who in this place set up a sundial, to cut and hack my day so wretchedly into small pieces!" -- Plautus, 200 BC
signature.asc
Description: Digital signature
