Re: Guile integration and UTF-8

[Paul Smith]

On Tue, 2013-09-24 at 23:11 +0200, Ludovic Courtès wrote:
> Hi,
> 
> Lars Ljung <address@hidden> skribis:
> 
> > 2) This simple example fails because of the UTF-8 string:
> >
> > all:
> >     @echo $(guile "Hej, världen!")
> >
> > The result is:
> >
> > ERROR: Unknown object: "Hej, världen!".
> 
> Make uses ‘scm_c_eval_string’, which expects to be passed a string in
> the current locale encoding.
> 
> So, if you have a UTF-8-encoded Makefile, but run ‘make’ with a
> ISO-8859-1 locale, the string will be incorrectly decoded.
> 
> In particular, ‘to-string-maybe’ in gmk-default.scm will consider that
> some characters are not printable, leading to the “Unknown object” error.
> 
> The fix would be to convert the string passed to $(guile ...) from a
> fixed encoding, rather than from the current locale encoding.  For
> instance:
> 
>   scm_eval_string (scm_from_utf8_string (argv[0]))
> 
> Better yet, use ‘scm_file_encoding’ to determine the encoding of a
> makefile based on the presence of -*- coding -*- cookies (info "(guile)
> Character Encoding of Source Files").

Thanks Ludo'.  I figured it was something like that.

I haven't made a formal definition, but I'm seriously leaning towards
declaring that makefiles are UTF-8, always, and not supporting other
encodings.  I'd like to discuss the possible downsides with others who
are more versed in i18n issues than I before I do this, and I won't do
it for this release, but I'm really expecting to do it (and document it
as such in the manual, etc.) shortly.