Re: undefined-behavior obstack.c:139

[Paul Eggert]

On 12/4/23 05:57, Bruno Haible wrote:

> In both cases, the ability to add NULL + 0 unifies code paths for
> empty arrays and non-empty arrays. Such unified code paths are desirable:
> they make the code look simpler.
> ...
> Paul, is the attached patch OK with you?

Thanks, but I'm of two minds about that sort of thing. On the one hand, 
they document the assumption that (char *) NULL + 0 == NULL. On the 
other, they are still clutter and this assumption is safe on all but 
pedantic platforms -- and we've long not worried about pedantic 
platforms (e.g., gcc -pedantic), for good reason.

Of course, the fact that you found these issues with clang's runtime 
checking does not mean that these are all the issues; I fully expect 
that there are others.

All in all, it may be better to leave the code alone.

PS. The name END_OF_ARRAY is not the best, since a null pointer does not 
point to an array, and even when the pointer is non-null the macro can 
be used to point elsewhere than to the end of an array. A better name 
would be something like PTR_ADD, since the first arg is the pointer, the 
second arg the amount to add.