Bad output when compressing empty array, with Compression::none()

[xNxExOx]

I think this test example is enough to demonstrate the issue: Rust Playground

You can uncomment dbg!(&compressed_table); to see the huge result, that can not be decompressed afterwards.

Priority low, I resolved the issue on my side by using flate2::Compression::default() for empty tables, which works fine.

This is really weird edge case, so no need to fix it any time soon, because not many people is compressing empty stuff, and even less does so with flate2::Compression::none().

[oyvindln]

Which version of miniz_oxide is it using? There was a bug introduced in miniz_oxide 0.8.3 that I fixed in 0.8.5 that caused corrupt output in this case so maybe rust playground hasn't updated yet. I was considering yanking the two versions though it seemed like an extreme edge case to happen in the wild though I guess if it does occur maybe I should as people can be slow to update.

You very rarely want to intentionally use no compression as zlib and other compressors already switch to it automatically if the data is not compressible (though there are probably edge cases where you would to intentionally not compress it for whatever reason). An "empty" file will be smaller with any compression level higher than none as the block header metadata for a deflate compressed block with static codes is smaller than for an uncompressed block.

[xNxExOx]

Well my reason for using none might be stupid, but I sometimes read the files in hex editors, and dealing with double, or even triple compressed files is annoying, but that is weird edge case on its own. It have nothing to do with real size. Also having it without compression make it clear when the size grows over few GB, which compression hides very well to few MB, because that part of data is too well compressible.

For me locally I have: miniz_oxide: version = "0.8.4" based on Cargo.lock. Even after updating to flate2 version = "1.1.0" from version = "1.0.35". So I am not sure, if that is the issue.

And finding out that one of the tables is empty was surprise for me, so that is huge edge case too.

[xNxExOx]

Even main branch is still on that version: https://github.com/rust-lang/flate2-rs/blob/main/Cargo.toml#L27

[oyvindln]

If you run cargo update it should update to the latest version.