[darcs-users] working around encoding problem

[Wolfgang Jeltsch]

Am Sonntag, 29. Oktober 2006 08:02 schrieb Eric Y. Kow:
> On Thu, Oct 26, 2006 at 17:16:21 +0200, Wolfgang Jeltsch wrote:
> > to my knowledge, patch names and long comments are still byte streams to
> > darcs.  So if one user uses Latin-1 and another UTF-8, characters outside
> > the ASCII range are stored in different encodings, depending on who
> > created the respective patch.  Is there a workaround for this?
>
> At the moment, no.  Is this a problem for UI issues, like how patches
> appear in darcs changes or being able to match on patch names, or is
> there a deeper more troubling issue?

The problem is that patches in a central repository will use different 
encodings for their names and long comments, depending on who created them.  
Most of the upcoming darcs users use ISO-8859-1 as their default encoding but 
I use UTF-8.

> > I already thought about setting up a Subversion-to-darcs proxy where
> > users can commit to a Subversion repository and every patch is
> > automatically converted to a darcs patch using a unique encoding for
> > patch names and long comments, and then applied to a specific darcs
> > repository.  Could this be realized with Taylor?  How?  Is there a
> > simpler approach?
>
> Perhaps the good people on the Tailor list might would know more.
> http://lists.zooko.com/mailman/listinfo/tailor

Meanwhile, I started to create an solution myself which is not as heavy as 
using such a Subversion-to-darcs proxy.  My approach is to create an 
executable called darcs.exe which checks for the presence of a --logfile 
option and changes the encoding of the file denoted by this option to UTF-8 
if such an option exists.  Finally, it calls the actual darcs binary using 
the arguments it was called with.  Since all the ISO-8859-1 users will use 
TortoiseDarcs and TortoiseDarcs seems to denote patch names and long comments 
only via logfiles, this should work for us.

The problem, I'm currently struggling with is that I cannot get MingGW's 
linker to properly link with libiconv. :-(

> Good luck,

Thank you. I'll probably need it…

Best wishes,
Wolfgang