bug/feature request: apply textconv in "git add -p" diff output

>
> Hi,
>
> I use git to manage bdf font files. It it quite cryptic so textconv
> for diff output makes it much easier to keep track of changes. The
> only problem is that "git add -p" does not seems to use textconv, so I
> have to run in parallel "git diff" and "git add -p", then add chunks
> accordingly. Can somebody add textconv support to "git add -p" please?
> I'm not so good at Perl to do the job.

> Hi,
> 
> I use git to manage bdf font files. It it quite cryptic so textconv
> for diff output makes it much easier to keep track of changes. The
> only problem is that "git add -p" does not seems to use textconv, so I
> have to run in parallel "git diff" and "git add -p", then add chunks
> accordingly. Can somebody add textconv support to "git add -p" please?
> I'm not so good at Perl to do the job.

> Nguyen Thai Ngoc Duy venit, vidit, dixit 20.06.2009 09:17:
>> Hi,
>>
>> I use git to manage bdf font files. It it quite cryptic so textconv
>> for diff output makes it much easier to keep track of changes. The
>> only problem is that "git add -p" does not seems to use textconv, so I
>> have to run in parallel "git diff" and "git add -p", then add chunks
>> accordingly. Can somebody add textconv support to "git add -p" please?
>> I'm not so good at Perl to do the job.
>
> I don't think this is possible: textconv filters are meant to produce
> easily readable output from diff, but in no way are they required to be
> injective mappings. So, textconv output cannot to be used in order to
> specify a patch uniquely.
>
> The appropriate approach is to use smudge/clean filters. They are
> mutually inverse. And, in fact, I just checked: add -p uses the cleaned
> version. So, if clean converts to "something textual" which gives you
> readable diffs and meaningful stats (and smudge converts to binary bdf)
> then that is the way to go.