================ @@ -215,3 +215,10 @@ namespace PR20735 { // fix-it:"{{.*}}":{[[@LINE-9]]:20-[[@LINE-9]]:20}:")" } } + +void consecutive_builtin_compare(int x, int y, int z) { + (void)(x < y < z); // expected-warning {{comparisons like 'X<=Y<=Z' don't have their mathematical meaning}} + (void)(x < y > z); // expected-warning {{comparisons like 'X<=Y<=Z' don't have their mathematical meaning}} + (void)(x < y <= z); // expected-warning {{comparisons like 'X<=Y<=Z' don't have their mathematical meaning}} + (void)(x <= y > z); // expected-warning {{comparisons like 'X<=Y<=Z' don't have their mathematical meaning}} ---------------- JOE1994 wrote:
(I'll refer to each of the 3 cases mentioned in the above comment as case1, case2, and case 3) I pushed a commit that adds 2 `no-warning` test lines for case1. It's unlikely for one to write `((x < y) >= z)` and believe it to have mathematical meaning of `(x < y) AND (y >= z)`. I limited the scope of this PR to diagnosing consecutive "built-in" comparisons (as per the PR title). Thus, the current implementation does not diagnose case 2. I don't have the capacity to further work on making clang diagnose case 2 but not case 3. I added a TODO note at the bottom of the test lines for now. FYI, gcc doesn't diagnose any of case1, case2, or case 3. https://github.com/llvm/llvm-project/pull/92200 _______________________________________________ cfe-commits mailing list cfe-commits@lists.llvm.org https://lists.llvm.org/cgi-bin/mailman/listinfo/cfe-commits