Another bogus "Conditions should not unconditionally evaluate to "TRUE" or to "FALSE"" warning.

[ktullavik]

When changing a variable with unary operators, the analyser don't seem to invalidate its
assumptions - and produces false warnings.

Example, both inner conditionals in this function give bogus warnings:

    void test() {
        int b = 2;
        if (b > 0) {
            b--;
            if (b == 0) {
                return;
            }
            if (b > 0) {
              return;
            }
        }
    }

Funny enough, the warning would be appropriate for the outer conditional.
Dart SDK version: 1.22.1

[bwilkerson]

@alexeieleusis

[mraleph]

@ktullavik b is equal to 1 after b-- so b == 0 evaluates to false and b > 0 evaluates to true unconditionally. what is bogus about the warning?

[alexeieleusis]

These cases are indeed not considered in the rule, there is a bug reported for it and a PR out dart-archive/linter#412 should land soon.

[alexeieleusis]

By the way, the rule does not try to infer the value of the variable, and should not warn about anything inside the outer if statement precisely because of the mutation to b.

[ktullavik]

@mraleph the warning is not dependant on the value of b.

[mraleph]

@ktullavik no matter which constant value you select for the initial value of b conditions like b == 0 and b > 0 would always be constant - because b value after b-- is known. They might be different depending on the constant value of b you selected, but they are nevertheless known in compile time.

Compare your example with

void test(bool x) {
  int b = x ? 2 : 1;
  if (b > 0) {
    b--;
    if (b == 0) {
        return;
    }
    if (b > 0) {
      return;
    }
  }
}

Here indeed conditions are not constant.

[ktullavik]

@mraleph
Yes sorry, I should have made b unknown in the example to avoid confusion.
But it's irrelevant. The wrong analyser assumption is caused by the outer conditional,
not by the value of b.

[srawlins]

This linter rule is removed.