[darcs-devel] [patch1628] windows fixes for 2.14.0

[Ganesh Sittampalam]

Ganesh Sittampalam <ganesh at earth.li> added the comment:

The latin9 regression seems more serious, though. I didn't think very hard about the semantics of what I was 
doing when I wrote the build fix patches, and I suspect the Darcs.Util.Encoding one is completely wrong - for 
Windows it just goes back to the old behaviour, and it seems like that mishandles things being passed on the 
command-line.

I played with trying to just get Windows to do the same thing as other OSes in Darcs.Util.Encoding and 
Darcs.Utils.External (i.e. set the foreign/locale encodings to the filesystem encoding, and use the knob-
based code) but I get this error:

> darcs.exe: darcs-encode: commitBuffer: invalid argument (invalid character)

I suspect that even the filesystem encoding doesn't roundtrip properly on Windows :-( particularly given this 
comment in the docs about using the W APIs instead: https://hackage.haskell.org/package/base-
4.10.1.0/docs/GHC-IO-Encoding.html#v:getFileSystemEncoding

I'll keep playing, but I'm a bit lost about how this should work.

__________________________________
Darcs bug tracker <bugs at darcs.net>
<http://bugs.darcs.net/patch1628>
__________________________________