bug#38: No coding system used for environment variables

[Chong Yidong]

> It seems there is no coding system applied to values of environment
> variables.

What's the current situation with this bug?  Jason's patch was reverted,
but nothing else seems to have been done after that.

Two objections were made to Jason's patch: (i) some coding systems are
not ready until some .elc files get loaded (relevant for special cases,
such as the EMACS_LOAD_PATH variable), and (ii) DECODE_FILE causes GC,
so variables such as `nm' in Fexpand_file_name may not point to valid
data after that.

If no elegant solution is forthcoming, I'd suggest simply documenting
(i) as a limitation, and dealing with (ii) by simply turning off GC in
the affected part of the function.

I noticed that the patch posted at

http://lists.gnu.org/archive/html/emacs-pretest-bug/2007-05/msg00115.html

has not been checked into the trunk either.