Re: [PATCH (GIT-GUI,GITK) 0/8] Encoding support in GUI

[Paul Mackerras <paulus@samba.org>]

Alexander Gavrilov writes:

> File encoding can be specified in the following ways:
> 
> 1) It defaults to the current locale encoding.
> 2) It can be overridden by setting the gui.encoding option.

I'm happy with providing a way to say what the default encoding of
files in the repository is, but I wonder why it is seen as a property
of the GUI.  Is it just that there is an existing "gui" section that
is convenient to use, or does git-gui already use gui.encoding (before
this patch series), or is there some other reason?

> 3) It can be further set on per-file basis by specifying
>    the 'encoding' attribute in gitattributes.

I haven't used .gitattributes before, but I would expect that the
.gitattributes files would be stored in the repository along with
everything else.  If that's the case, then for gitk at least there is
the question of which version of a given .gitattributes file one
should use when viewing the tree for a commit which isn't the
currently checked-out commit - do you use the version from that tree,
or the version in the working directory?  We seem to be using the
latter at present, and caching the results.  Is there a philosophical
reason to do that, other than speed?  (Also it seems that we won't
notice if the user changes .gitattributes after we've looked at it, or
if they create one after we've looked for one and not found it.)

> Since git apparently cannot work with filenames in non-locale
> encodings anyway, I did not try to do anything about it apart
> from fixing some obvious bugs.

For Linux, filenames are sequences of non-null bytes, so using
binary encoding to read them in Tcl sounds about right.

> There are also some bugs in handling of commit encodings in gitk,
> but they are out of the scope of this series.

I'm interested to hear what they are.

Paul.