[RFC] scanf: add non-allocating bounded string input via %!s / %![...] / %!c (size before pointer)

[Joseph Myers]

On Fri, 17 Oct 2025, Yair Lenga wrote:

> * No other format decorations are allowed with '!':
>   - **no** numeric field width digits,
>   - **no** length modifiers (h, hh, l, ll, L, j, z, t),
>   - **no** positional "$".
>   If present, the conversion is a format error (treated as a match failure).

I'm concerned about the non-orthogonal design being proposed, as well as 
the problems for i18n support.

For printf width and precision, "*" can also be given as "*N$" to take the 
Nth argument.  This is necessary to support %N$ argument reordering for 
i18n, since mixing numbered and positional arguments in the same format 
string is unsupported.

The same principle should also apply for anything taking a size argument 
for scanf: it should be able to have N$ specified so it can be used in 
strings with %N$.

The existing "*" for printf takes an int.  There's a proposal for C2y 
(N3691) to allow a length modifier for precision and add the option of '^' 
instead of '*' for unsigned precision (the proposal only deals with 
precision, not width).  The proposal doesn't deal with "*N$" since that's 
a POSIX feature not an ISO C one, but one might reasonably expect to be 
able to use "*N$" and "^N$" together with the new support for length 
modifiers.

Now, "*" already has an existing incompatible meaning for scanf so can't 
be used for a new feature here.  But in the other ways the proposed 
feature could be more consistent with printf - so supporting a positional 
modifier, and defaulting to int with potential support for length 
modifiers if such a feature is accepted for printf in C2y in future.

-- 
Joseph S. Myers
josmyers@redhat.com