[Bug-ddrescue] Re: Using ddrescue on some really bad floppies

[Ian Oliver]

In article <address@hidden>, Ian Oliver wrote:
> I'm about to roll up my sleeves and write some Python to parse the 
> ddrescue log and then try to use fdrawcmd to pull off the sectors while 
> ignoring CRC errors. I'm not desperately hopeful, but assuming the 
> controller can find the headers, perhaps I can pull the damaged body 
> off several times then try and use semi-psychic techniques to patch up 
> some data.

Just for closure, here is what I did.

1) Pull multiple copies of data off disks using a "Catweasel" card, which 
does *very* low level access.
2) Write python parser to root through the data dumps and find sector 
headers and data blocks. It was at this stage I discovered the sector 1 of 
track 1 to 35 said it was one track further on that it was. Nice.
3) Once I had all the sectors that ddrescue failed to get (mainly those 
with duff track numbers), use ddrescue to pull these from output of parse 
process and into the image.
4) Run obsolete proprietary backup utility in a DOS machine under 
virtualbox, point it at the "floppies" one at a time, see long-lost source 
code for old game (Carrier Command) being written the vdi file.
5) Mount vdi file in Linux, grab data, and put it SOMEWHERE SAFE!

I'm still one sector short on each of two disks, but can keep scraping 
away with Catweasel. Oddly, the data is identical in all the copies, but 
the crc is wrong. I plan to examine the clock bits to get a hint at which 
byte(s) is/are wrong, and then find values that fix the crc. Wish me luck!

Ian