https://github.com/ocaml/opam-repository/pull/13118#pullrequestreview-183576751

Suggested-By: @mseri on Github

Ugh, experiments show that MS will parse arbitrarily short hex encoded
IAs which in turn revealed miscalculation in our sloppy parsing mode.

Increment major version due to API breakage.

In MS mode we need for much shorter strings to pass through to the
hex parser.

Provide immediate feedback on stdout when running in an interactive
terminal the assumption being we’re dealing with a human.

Add a conformance handler “Con_MS” to achieve a behavior that
mimicks that of MS’s implementation bug-for-bug.

Aspects of reading and formatting governd by the conformance:

    - Validation of ident auths,
    - validation of subauths,
    - validation of leading zeros in decimal numbers,
    - zero-padding of hex numbers.

Offload the test for member count and size constraints of sas and ia,
respectively, into a separate validation module. This is a preparatory
step toward adding less rigid parsing modes.

Both the constructor “Sid.create” and the string format parser must
reject inputs lacking a subauthorities array of at least size one. Since
the array is no longer optional, reorder the the constructor arguments
to match the data representation.

It is still possible to create SIDs without subauthorities via the
“Sid.create_unsafe” constructor. Also, the packet representation will
happily accept them because their definition (as well as that that of
the identical RPC version) does not specify a minimum count.

This is all rather ambiguous and exacerbated by the fact that [MS-DTYP]
happily specifies an invalid SID “S-1-5” as the “NT_AUTHORITY”. However,
both the grammar and the Win API “ConvertStringSidToSidA()” function
reject SA-less inputs as invalid, so we should too.