unsafeFromUTF8BS in D.FieldGrammar.Parsec is wrong

[phadej]

The bytestring can in fact be invalid UTF8, previously when parsing to String we did conversion using fromUTF8BS which leniently fixed the text using replacement character.

I'm not sure what to do here. Eq is broken but can be worked around in tests (I'll push a commit soon). Conversion of ShortText to String introduces replacement characters as well.

Performance wise one would like to do nothing, we don't compare free text fields for equality inside Cabal.

[phadej]

Looks like there is only single invalid UTF8 file on Hackage. I'll add it to quirks.

The local file issue still persists. I will close this issue however, I cannot think how it could it break people.

[phadej]

Fixed in 8a569d8