rename only repair mode (quickly skip non matching files)

[sjpotter]

I currently ran into a situation where I accidentally renamed 1000 files (about 270GB in total) of files protected by multiple par2 sets (but all mixed together now).

It seems the only way to recover it is to iterate over every par2 file and pass in all the renamed files to the cmdline so it can see what can match.

However, it tries to do this to find blocks in all the files, so it scans them very slowly to do this.

I'm wondering if there would be better to have a "rename only" mode, where we make the assumption that the file really has no errors besides being renamed, so as soon as we see the start of the file doesn't match any of the files we expect, we can ignore it and move on, thereby saving huge amounts of IO.

an extra to this that would apply to regular repair as well would be, once the "needed extra blocks" drop to 0 (i.e. we have all the data needed and dont need to actual recover data, just rename the files we have) we can stop processing the extra files passed in and just go to rename mode if required.

thoughts?

[sjpotter]

my quick and dirty thought for this is that in ScanDataFile() when we realize it's only a partial match, we return that its no match instead when in this mode.

perhaps there's a better way?