Feature request: reuse compressed blocks for incremental updates

[TechnoPorg]

I'm currently working on a project which mounts a squashfs from a live CD image with a writeable overlay, makes some changes, then packs it back up. Doing so is the slowest part of the build process, and it would be great to shave off a bit more time if possible.

I realize that fully incremental updates of a squashfs are undesirable (as per your detailed answer here: #174 (comment)). However, there are many files in the original squashfs larger than the block size, and if I understand correctly, the blocks that aren't part of a fragment won't change if the file contents don't change. Would it be possible to re-use the compressed blocks by reading them from the original, rather than compressing them again, and then store them as usual?

I also just wanted to say that the upgrade to mksquashfs 4.7.2 has already sped things up noticeably, and I really appreciate your continued hard work on this project!

[plougher]

It should be possible to do that. But until now I have not been asked to do it, and as such it is has never been a high priority.

I'll do some thinking to see how difficult it is to do, especially how much of the existing code can be re-used. Generating a new filesystem from an original is much easier than modifying a pre-existing filesystem in-place.

[TechnoPorg]

Thank you!

And just to clarify, I'm not proposing an in-place modification of the existing filesystem, but just that it would read the compressed blocks from it, and then immediately store them in the new filesystem. You would know best whether the former would be better, though.