Patch used to be small but isn't anymore?

[UltimateWalrus]

Hello,

I'm trying to create a patch that will convert from one version of a classic game (Little Big Adventure 2) to another. Primarily the major space-hog is a CD image, LBA2.GOG (both versions over 600 MB). The GOG format is actually the same as a BIN with an accompanying CUE file for describing the layout of data on a CD.

I would expect there to be a lot of duplicate data between the two versions of the file. Granted it may be in completely different places and difficult to find. But I don't think it would be compressed.

Currently the Patch Maker algorithm seems to be finding no similarities between the files that it can use to save space. I would simply accept that the algorithm can't figure out any similarities. The only thing is, at one point in the past I somehow built a test version of my patch that miraculously came out to only 8 MB (i.e., at one point something "clicked" and Patch Maker actually found many similarities between the files).

Is there some way to coax Patch Maker into seeing the similarities between these two files again, maybe something that was present before but then got changed somehow? I don't think I accidentally left an identical file when I built the smaller patch (I tested this just to make sure and it would have been less than 4MB).

If it would be helpful I can try to send the files somehow (over a GB of data) but don't want to go through the effort unless you think it will be useful.

Thanks,
Sebastian

[Yves]

2 possible issues:

- If the data are compressed, then a little change will change the entire part that contains the compressed data. For example if you change a single pixel in an image and zip it this will usually affect the entire compressed file.

- IIRC PM cannot find similarities that are separated by more than 256 or 512 Kb of data, it easily loses the synchronisation between big files.

[UltimateWalrus]

Thanks for your response. The big file must have accidentally been left out of the small patch I built. I'll just have to accept my patch will be 600MB.

Out of curiosity and stubbornness I wrote my own patching algorithm. The problem with disc image files is not that they are compressed, but that they are chopped into 2048-byte sectors, which could be rearranged completely in the newer version (like a jigsaw puzzle). I ran my algorithm for over a day and ultimately it found 200MB of similarity between the files, but something was bugged in my actual patcher code as the file checksums did not match after. I was hoping for more than 200MB savings anyway.

Interestingly, my patcher seems to work well with MMF2 EXEs. I successfully converted from one version of a 16MB game to another with a 35KB patch. Maybe it will be useful for something someday.