Re: LC_ALL and UTF-8 directory names

[Akim Demaille]

Hi Ben,

Le 5 oct. 2011 à 18:40, Ben Pfaff a écrit :

> Akim Demaille <address@hidden> writes:
> 
>> The JVM is quite happy with that.  Unless you make the name
>> "invalid" by setting LC_ALL to C for instance, which Autoconf
>> does.
> 
> Glib supports a G_FILENAME_ENCODING environment variable that
> directly controls the encoding used for file names.  Does the JVM
> have any equivalent?

Thanks for the suggestion.  I have tried to find something on the Internet, but 
I saw nothing conclusive.  Some people seem to believe -Dfile.encoding=UTF-8 
might help (but actually it's about the encoding for file I/O), but others 
point out that sun.jnu.encoding should do it 
(http://happygiraffe.net/blog/2009/09/24/java-platform-encoding/).  
Unfortunately it does not work here, and some are angry at the JVM 
(http://bugs.sun.com/bugdatabase/view%5Fbug.do?bug%5Fid=4733494).  They advice 
to define LC_ALL properly 
(http://stackoverflow.com/questions/5663709/how-to-fix-java-when-if-refused-to-open-a-file-with-special-charater-in-filename).

Apart from moving away from this directory with non-ASCII characters, the only 
way out I see currently is really to restore the locales when invoking the JVM. 
 But that requires getting the user's locale, which is not something that 
Autoconf 2.68 allows to do:

$ grep LC_ configure
LC_ALL=C
export LC_ALL
LC_ALL=C
export LC_ALL
$

configure should probably save the previous locale values before overriding 
them all, that might be help on occasions, such as this one.