Currently-harmless UB in stack_reset() and stack_reallocate()

[nicowilliams]

Reported here.

clang w/ -fsanitize=undefied reports:

../src/exec_stack.h:64:32: runtime error: applying zero offset to null pointer
SUMMARY: UndefinedBehaviorSanitizer: undefined-behavior ../src/exec_stack.h:64:32 in
../src/exec_stack.h:83:36: runtime error: applying zero offset to null pointer
SUMMARY: UndefinedBehaviorSanitizer: undefined-behavior ../src/exec_stack.h:83:36 in

These are cases doing pointer arithmetic with NULL pointers and then yield NULL pointers. If the compiler were to try to optimize these functions away due to this, then that would cause jq to break immediately -- the tests certainly wouldn't pass, especially with valgrind. Therefore it is safe to say that for all the tested builds of jq this bit of UB is not currently a problem.

[emanuele6]

@nicowilliams I think this is the same bug reported in and fixed by #2108

[nicowilliams]

@nicowilliams I think this is the same bug reported in and fixed by #2108

Yes. Though #2108 would need to be rebased since it has other fixes that are no longer needed.

[emanuele6]

Can we just fix this? UBSAN occasionally hits this e.g. when running jq --build-configuration, and it's annoying.

[nicowilliams]

Yes, we can. I meant only that this didn't cause problems yet, so we didn't have to fix it for 1.7.