human interface change?!? [Re: xstrtol.h

[Jim Meyering]

Bruno Haible <address@hidden> wrote:
...
> I am committing your patch to xstrtol.h, with updated variable names and
> comments. Also, a corresponding change to the 'human' module. Find attached
> the corresponding coreutils change (untested but obvious).
>
> 2007-07-25  Jim Meyering  <address@hidden>
>             Bruno Haible  <address@hidden>
...
> *** NEWS      14 Jul 2007 16:25:53 -0000      1.24
>   Date        Modules         Changes
>
> + 2007-07-26  human           The function human_options takes an additional
> +                             third argument, the option that introduced the
> +                             argument that was passed to a xstrto* function.

Your committing such a change comes as a big surprise.
First of all, I said I would check in my change to xstrtol.h.
Next time you're impatient to change something I own, ping me instead.

More importantly, don't you think you should have waited for approval
from Paul (and discussion on this list) before making an _interface change_
to the "human" module?  Especially when I already expressed reluctance
to change that part?

My concern is that uses of human may well be in contexts where there is
*no* option/argument involved.  Hence you've now introduced an artificial
constraint in making its diagnostic speak of an "argument".  In addition,
you're requiring all callers (I suspect there are many, in other packages)
to adapt to the interface change -- with no prior discussion.

As you can see, I disapprove of the human.h change (at least now, since
I've seen no discussion at all), so also object to your using my name
in the corresponding ChangeLog entry.

Finally, regarding your proposed coreutils patch, do you really think
that passing a NULL "option string" into human_options is acceptable?

Please revert those changes.