dust shows IEC values with SI prefixes

[Zorbatron]

The manual shows these options for --output-format:

• si
• b
• k - t for KiB - TiB (IEC)
• kb - tb for KB - TB (SI)

To me, this implies that it defaults to SI prefixes, and the output using M or G suggests the same. However the reported sizes are actually using IEC prefixes and are "larger" when specifying -o si.

Testing with these files:

$ fallocate -l 1000000000 1GB.bin
$ fallocate -l 1G 1GiB.bin
$ fallocate -l 1000000 1MB.bin
$ fallocate -l 1M 1MiB.bin

eza shows the correct units:

$ eza -l --no-permissions --no-user --no-time (default is SI)
1.0G 1GB.bin
1.1G 1GiB.bin
1.0M 1MB.bin
1.0M 1MiB.bin

$ eza -l --no-permissions --no-user --no-time -b (for IEC prefixes)
954Mi 1GB.bin
1.0Gi 1GiB.bin
977Ki 1MB.bin
1.0Mi 1MiB.bin

$ eza -l --no-permissions --no-user --no-time -B (for plain byte count)
1,000,000,000 1GB.bin
1,073,741,824 1GiB.bin
    1,000,000 1MB.bin
    1,048,576 1MiB.bin

Whereas dust doesn't differentiate between SI and IEC:

$ dust -bo si
1.0M   ┌── 1MB.bin
1.0M   ├── 1MiB.bin
1.0G   ├── 1GB.bin
1.1G   ├── 1GiB.bin
2.1G ┌─┴ .

$ dust -b
980K   ┌── 1MB.bin
1.0M   ├── 1MiB.bin
953M   ├── 1GB.bin
1.0G   ├── 1GiB.bin
1.9G ┌─┴ .

$ dust -bo b
   1003520B   ┌── 1MB.bin
   1048576B   ├── 1MiB.bin
1000001536B   ├── 1GB.bin
1073741824B   ├── 1GiB.bin
2075795456B ┌─┴ .

(Idk why dust reads a wrong size for the SI files, eza, GNU ls, and GNU du don't do this)

When printing the sizes with IEC prefixes, it should be Ki/Mi/Gi/Ti. The man page should also mention that it defaults to IEC prefixes.