[PATCH 12/15] stdio-common: Fix NaN input data for scanf input specifier tests [BZ #32857]

[Maciej W. Rozycki]

On Tue, 6 May 2025, Andreas Schwab wrote:

> > -%la:-nan(09A_Zaz):1:13:-nan:
> > +%la:-nan(09A_Zaz):1:13:-nan(09A_Zaz):
> 
> Since 09A_Zaz is not a valid number, this is actually read as the
> default NaN, not as a NaN with a payload, and should be verified as
> such.

 It's an implementation detail we don't need or indeed want to rely on.

 The `n-char-sequence' token does not have to be a number and neither ISO 
C nor POSIX say how to interpret non-numeric or indeed any values.  All 
the tests are meant to verify or rely on is that a given `n-char-sequence' 
value is interpreted the same way when handed as input to `scanf', etc. as 
it would when passed as an argument to `nan', etc.

 So I maintain it is correct test input given here.  If you'd like various 
`n-char-sequence' values verified for specific payload patterns produced, 
then it will have to be a separate test case.

  Maciej