Re: Regression with printf 8-bit format macros

[Craig Howland <howland at LGSInnovations dot com>]

On 01/04/2016 02:16 PM, Ambroz Bizjak wrote:
> Hi Craig, others,
>
> The quoted part of the standard does explain how "hh" shall behave,
> but I do not see any implication that PRI*N should use hh or generally
> that the described conversion (after promotion) is required for PRI*N.
> (note: I suspect you meant to use PRIu8 in the example not PRId8
> because the conversion of 1000 to int8_t is an integer overflow.)
Yes, it is implied, by the "suitable for use within the format argument of a 
formatted input/output function when converting the corresponding integer type" 
clause.  Not only that, the very existence of the macros implies it.  Why would 
you bother with ""%"PRId*" rather than just using "%d"?  You use the macro to 
supply the correct modifier that you would use if you knew the type (and that 
the type were invariant).  If the macros never supplied modifiers, then they'd 
be useless.
I copied the example directly from the opening email, which used d. I should 
have said that -24 should be printed--the signed interpretation of 0xE8--rather 
than 232.  (I used an unsigned method of giving myself the decimal initially.  
1000=0x3E8, which will be passed as that (assuming ints are >=16 bits), then the 
cast to signed char truncates to 0xE8, is either 232 or -24.)
> About the %n issue, it has no relation to the PRI* macros. The PRI*
> macros cannot possibly used together with "n", because each PRI* macro
> must end with a conversion specifier, which is not "n", and there can
> only be one conversion specifier in a conversion. For example,
> "%n"PRId8 and "%"PRId8"n" could become "%nd" and "%dn" respectively
> (the first one writes out the position then prints a literal d, the
> second prints a integer and a literal n). Notice that the standard
> does not define any "PRIn..." macro.
My bad on that, I failed to check the list properly for PRIn... . (I guess I 
shouldn't rush so much next time.  Sorry for the noise.)
> Best regards,
> Ambroz