rsmith added inline comments.
================
Comment at: clang/test/Sema/pointer-addition.cpp:5-6
+ char *f = (char*)0;
+ f = (char*)((char*)0 - f); // expected-warning {{performing pointer
arithmetic on a null pointer has undefined behavior}}
+ f = (char*)(f - (char*)0); // expected-warning {{performing pointer
arithmetic on a null pointer has undefined behavior}}
+ f = (char*)((char*)0 - (char*)0); // valid in C++
----------------
These two warnings are wrong -- their claim about these expressions having
undefined behavior is incorrect. We can't prove that `f` is not null (and in
fact it is null here), so we should not be producing a warning that says the
code has undefined behavior. If you want to warn in the cases where you can
prove the other pointer is non-null, and say that that case has undefined
behavior, that seems fine, but please fix the diagnostic message to be
technically correct (eg, "computing difference of a null pointer and a non-null
pointer has undefined behavior").
Perhaps a better approach would be to use the same logic to decide whether to
warn in C and C++, but produce different warning text. For example, you could
say "[...] has undefined behavior" in C, but in C++ just say "warning:
performing pointer arithmetic on a null pointer" without making
potentially-inaccurate claims about UB? The code is still *suspicious* in C++
even if it's not UB.
Repository:
rG LLVM Github Monorepo
CHANGES SINCE LAST ACTION
https://reviews.llvm.org/D98798/new/
https://reviews.llvm.org/D98798
_______________________________________________
cfe-commits mailing list
cfe-commits@lists.llvm.org
https://lists.llvm.org/cgi-bin/mailman/listinfo/cfe-commits
