nhivanye on 14/3/2006 at 09:55

---

Shadowspawn,

I'll start looking at matlib stuff. Unfortunately I don't think I'll finish anything before Oblivion arrives next week :D which will absorb a lot of my spare time for a week or two at least :eek:

Shadowspawn on 20/3/2006 at 02:16

---

I managed a few more hours of work on this, and there's good news and bad news.

1.) Good news! Lots and Lots of the matching data type / size entries are EXACTLY the same, and can just be read from the other stock .IBT files. So we're good there.
2.) Other entries, although the same size, have small (1-4) byte differences. This makes it a tiny bit harder.

My current thought is to zero out the data portion of matching entries and fill them in from the stock IBT. I'm toying with the idea of zeroing out all but the differences, and not overwriting them during the stock refill. (there's problems, what if the correct value is 0, I don't want to overwrite it).

I figure ZIP will crunch down any region of 0s to almost nothing, and it makes inflating the CBT that much easier. I'm still a few hours / days away from a test version - stay tuned.

Crispy on 21/3/2006 at 06:00

---

Awesome.


You could keep a list of locations in your CBT (I'm assuming that stands for Compressed IBT) where the correct value is 0 but your program will overwrite it with a wrong value from the stock IBT. So then you inflate the CBT using the method you mentioned, and then you go through your zeroes-list and write 0 to each location. Obviously there are some optimisations to be made to this algorithm (like using the zeroes-list while inflating the CBT instead of afterwards, to avoid massive seek times when you're writing to widely-spaced locations in the newly-inflated IBT).

Might be a little inefficient, but I can't think of anything better off the top of my head.

Oh, and 0 is probably a poor choice for what I'm describing, since there are likely to be many 0 bytes in an IBT. :) Perhaps 254 or 255 or so would be better.

epithumia on 22/3/2006 at 00:53

---

You could look at xdelta ((http://xdelta.org)).

Shadowspawn on 22/3/2006 at 03:48

---

Hopefully I'll have the compressor finished this week, I'll have to wait to try the inflator. At least we can see how much it's likely to save us.

I'll be on vacation next week in Aruba - a much needed break. I'll finish the inflator when I get back. I think this will work out pretty well, the bulk of the compression will be the textures, which won't compress well, and remain constant (more or less) from mission to mission.

Hey, any taffers live in Aruba? I'm buying the first round of drinks!