Re: Bash printf should diagnose integer overflow

[Paul Eggert]

On 3/13/24 11:13, Chet Ramey wrote:

> Thanks for the report. The most appropriate fix for this particular issue
> is to display an error message if printf returns < 0, instead of
> suppressing it unless the -v option has been set.

Oh, good point. This simplifies things a bit, though Bash still needs to 
do its own overflow checking for cases like "printf '%2147483648q' ''" 
and "printf '%*s' 2147483648 ''" when the Bash code itself is parsing 
the integer, rather than relying on printf(3) to do it.

Revised patchset attached. The first patch uses the fix you suggested; 
the remaining patches are similar to what I sent earlier, except the 
last one is simplified since it doesn't need to worry about inline width 
and precision when printf will do the checking. These patches are 
relative to Bash devel commit bf944fe91ffa97743ad86f6db6f3b84c78207a78 
dated today at 09:33:32 -0400.

0001-printf-now-diagnoses-underlying-printf-failure.patch
Description: Text Data

0002-maint-add-support-for-C23-style-stdckdint.h.patch
Description: Text Data

0003-Coalesce-multiple-PRIdMAX-definitions.patch
Description: Text Data

0004-Fix-problems-with-large-brace-expansions.patch
Description: Text Data

0005-printf-now-diagnoses-more-width-prec-overflow.patch
Description: Text Data