Filename encoding interoperability between Linux

[Manuel Jacob]

I didn’t find an updated version of the fixutf8 extension.  I didn’t try 
whether it works with current Mercurial versions.  It seems like the 
approach taken by this extension is brittle.

Superficially, it seems to me like the amount of work to finish the 
relevant parts of the Windows UTF-8 plan and the amount of work to 
update the fixutf8 extension is comparable.

As a TortoiseHG developer, do you see any problem in supporting the 
approach described in the Windows UTF-8 plan in TortoiseHG (both in the 
application and in the shell extension)?

On 2019-08-05 15:13, Yuya Nishihara wrote:
> On Mon, 05 Aug 2019 02:19:04 +0200, Manuel Jacob wrote:
>> So far, all systems from which we accessed our Mercurial repositories
>> have been Linux machines, configured with a UTF-8 locale.
>> 
>> We want to give some Windows machines access to some of our
>> repositories.  But now the problem is that if a file with a non-ASCII
>> filename is committed on Windows, the filename is not properly shown 
>> on
>> Linux and vice versa.
>> 
>> I understand that the problem is that Mercurial doesn't interpret the
>> encoding of the file name.  But it's unclear how to solve the problem.
>> The Wiki was not very helpful regarding this.  The information ranges
>> from "too bad, Mercurial won't help you here" to "we have a plan on 
>> how
>> to fix the problem", but because the idea is originally from 2011, 
>> it's
>> unclear whether or not it was implemented in the meantime.
> 
> Windows UTF-8 support is not implemented yet.
> 
> https://www.mercurial-scm.org/wiki/WindowsUTF8Plan
> 
>> What would you recommend to solve this problem?  The Windows clients
>> access the repositories with TortoiseHG.  Ideally, in the repository, 
>> we
>> would like to have everything encoded in UTF-8.
> 
> I heard fixutf8 extension worked at some point, but I don't know if 
> it's
> compatible with recent Mercurial versions.
> 
> https://www.mercurial-scm.org/wiki/FixUtf8Extension