Re: [Monotone-devel] Problem with recursive add files to workspace

[Nathaniel Smith]

On Sat, Apr 28, 2007 at 12:18:04AM +0200, Thomas Keller wrote:
> Hi Maxim!
> 
> > If use curillic file name recursive addition is used does not work. 
> > If to specify a full path to a file then it it is added. 
> > If there is a cyrillic name of the directory that it in general it is
> > impossible to add in workspace. 
> > 
> > Under Linux all works normally
> 
> Unfortunately monotone's support wrt encodings under some filesystems is
> still very incomplete. The bug you mention here is a known bug, however
> one which is very nasty to fix properly, basically because there are so
> many things to look for. There exists a wiki page,
> http://www.venge.net/mtn-wiki/FileSystemIssues, which should list all
> major issues, however nobody could actually find time and resources to
> get this done right yet.
> 
> Sorry to say, but you're only safe with US-ASCII for now... :(

Well... if you use OS X, or a utf-8 locale on Linux/Solaris/BSD/...,
then you are also totally safe.  (Except for that one quirk with OS
X's unicode canonicalization, but that's much more minor than the
basic issue here.)  And if your locale is at least set correctly
(which apparently on some Windows machines it isn't?), then everything
except add -R, ls unknown, and ls ignored should work.  (I might have
forgotten a few commands, but basically it's just a small set.)

Also, fixing this is not nearly as hard as one might think from the
length of that page :-), though it will involve touching a fair amount
of source code -- basically we just need to be more careful about
tracking the character set that different strings in monotone are.
Doing the correct conversion is trivial and basically solved, just we
don't have the type system set up to catch us when we forgot to use
the existing machinery, so we re-introduce the same trivial bugs
repeatedly.

(There are also various Hard issues that touch on charsets, a lot of
which are mixed in to the discussion on that wiki page, but they can
mostly be separated out.)

-- Nathaniel

-- 
Electrons find their paths in subtle ways.