Re: Fwd: Re: fmtcheck() query

[uwe%stderr.spb.ru@localhost (Valery Ushakov)]

Christos Zoulas <christos%zoulas.com@localhost> wrote:

> | Yes, I'm so sorry for this breakage of file(1). It clearly indicates
> | possible risks of changing the behavior of fmtcheck(3).
>
> Thank you very much. I think that adding a new function is preferrable
> in this case because we are adding significant new functionality (by the
> positional arguments handling). We should leave the old fmtcheck(3) alone
> (so that existing code does not break) and just document its deficiencies
> and say to use fmtmatch(3) instead.

If the new function provides a flag that controls whether unused
trailing ares are ok, then fmtcheck() can be just a trivial wrapper.

One thing I sort of never figured out properly is convenient error
reporting.  My current code uses a fixed-size buffer in the
user-supplied struct, but asprintf(3) may be a better option.

I think error messages are important here, as visually matching the
rejected format and the template is a pain and is kidna stupid anwyay
since the function already knows exactly what's wrong.

-uwe