@silver18: Well, there were some upstream changes (you might want to check if some options changed behavior), sure, but mostly I just configured the build differently, it's now a bit less versatile, but much more Kindle friendly
. [Switched from quantum-depth 16 to 8, which means we lose precision if we ever have to process 16bit-depth PNG files, but that's not a common sight, at least for what we're doing, and used the iPhone codepath for the dithering algorithm. Long story short, for the rescale+dither workflow, I went from 167MB of total allocs w/ a peak at 64MB to 30MB of total allocs w/ a peak at 4MB, and it's probably around 10-20% faster.]