Re: [PATCH] gdb: add UTF16/UTF32 target charsets in phony_iconv

[Tom Tromey <tom@tromey.com>]

[ UTF-32 as intermediate encoding ]

Patrick> I nevertheless don't have any idea what is the amount of work required
Patrick> to change this.

Tom> I did most of it already.

After digging deeper into this, I think it can't work.  iswprint can
take the locale into account, and without this, we can end up in a
situation where the host-wide-char (or in this case utf-32) conversion
to the host multibyte charset ends up having to emit unnecessary
escapes.

For example in wchar.exp, \242 ends up being emitted as '\242\0\0\0',
because the \242 can't be converted to ASCII.  I think this happens
because gdb_iswprint in the utf-32 formulation disagrees with the iconv
conversion.

However, I think there's a better way to fix all this.  It's very simple
and offhand I don't know why I didn't think of it before... the use of
wchar_t depends on knowing the encoding of wchar_t -- and I think we do
know this on mingw, as it's a form of UTF-16.  So, maybe all that's
needed is a host-is-windows check in gdb_wchar.h, combined with a
suitable definition of INTERMEDIATE_ENCODING.

Can you try this?  Or if you'd prefer, I can send a patch for you to try.

thanks,
Tom