Re: uninorm/composition.c:75:22: runtime error

[Paul Eggert]

On 3/4/22 05:58, Simon Josefsson via Gnulib discussion list wrote:
> Below that is a patch for another UBSAN complaint about a NULL + 0
> operation -- I recall that being discussed before too, but don't recall
> the conclusion.

(char *)0 + 0 is undefined behavior, so clang's complaint about it is 
more justifiable than its complaint about assigning an unsigned char 
value to a char. But even with (char *)0 + 0 my feeling was that Clang 
is wrong, as adding 0 to (char *)0 works just fine on all platforms of 
practical interest; so we're better off overall by asking people to use 
-fno-sanitize=pointer-overflow when compiling Gnulib, if they also 
compile with -fsanitize=undefined. See the thread starting here:

https://lists.gnu.org/r/bug-gnulib/2021-10/msg00053.html

and ending here:

https://lists.gnu.org/r/bug-gnulib/2022-02/msg00049.html