Previously this used a quoted hex string; now it's a number with 0x prefix, just like the .octa directive.
Details
Diff Detail
Event Timeline
Especially if there is prior art this seems fine to me, especially if we keep the leading zeros.
The leading zeroes look weird to me, but I think in practice it's not going to be an issue anyhow so I think this is good to throw over the wall at the binutils list if you'd like?
Going to LGTM the patch for now as I think the syntax is fine as well.
The leading zeros are because an MD5 hash is defined to be a 128-bit value. MD5Result::digest() returns the full 32-character hex string. I didn't see any value in trying to suppress leading zeros, it's slow and more likely to be confusing than helpful.
The parsing should automatically zero-extend a short value, and as it's identical to .octa parsing (see test\MC\AsmParser\directive_values.s) I didn't bother with one for MD5.