Re: [PATCH] gitk: Fix display of copyright symbol in non-utf-8 locales

From: Jonathan Nieder <jrnieder@gmail.com>
Date: Mar 22, 2010, 23:45 UTC
Message-ID: <20100322234517.GB2168@progeny.tock>
In-Reply-To: <87pr2wc6bo.fsf@fox.patthoyts.tk>

Pat Thoyts wrote:

> Jonathan Nieder <jrnieder@gmail.com> writes:

Show 7 quoted lines

>> --- a/gitk
>> +++ b/gitk
>> @@ -1,6 +1,6 @@
>>  #!/bin/sh
>>  # Tcl ignores the next line -*- tcl -*- \
>> -exec wish "$0" -- "$@"
>> +exec wish -encoding utf-8 "$0" -- "$@"

[...]

> I would much rather see any fault in the makefile fixed than use
> source -encoding in all the various places this script might get
> loaded from.

Good point: git-gui.sh does not use ‘source’ but runs the moral
equivalent of
 
 set wish [info executable]
 eval exec $wish gitk $revs "--" "--"

to avoid starting a shell. Is it really so expensive to start a shell? This means that, for example, if some insane person were to rewrite gitk in C, then git gui would have to be adjusted before the result could be used.

Show 7 quoted lines

> THere are two characters in this file that are not ASCII
> and one of them is in a comment. We could even just use (c) to
> simplify this. However, \u00a9 is the correct unicode escape sequence
> for tcl to get a copyright character into the script in a way that is
> independent of the encoding. The escape sequence should be passed
> through unchanged in any translations - tcl wants to see exactly
> \u00a9 and everyone is using a copyright glyph here.

So we should fix the Makefile. The problem is that everything Tcl-related seems to assume files it sees will be in either the current locale’s encoding or ISO-8859-1 unless told otherwise, as far as I can tell. The entire gitk source tree, on the other hand, is in UTF-8.

Ideas? Jonathan