`!constant` in boolean expression not reported (function not called error)

Bug Report

It is a common source of bugs that developers fail to call functions in boolean expressions. This, among other reasons, the compiler checks for constant boolean expressions. However, there seems to be a hole in the existing checks. TSC currently allows “!fn” where “fn” and “(!fn) == true” are not allowed. It would be beneficial to check this kind of expression as well.

function fn() { return 'abc'; }

if (fn) {}  // ERROR: This condition will always ...
if (!fn) {} // not reported
if (fn == true) {} // ERROR: This condition will always ...
if ((!fn) == true) {} // ERROR: This condition will always ...

🔎 Search Terms

function “condition always true”

function constant boolean expression

🕗 Version & Regression Information

unknown, presumably since the beginning of time, at least TS 3.3, related improvement in TS 3.7

• This is the behavior in every version I tried, and I reviewed the FAQ for entries about Truthy expressions

⏯ Playground Link

Playground link with relevant code

https://www.typescriptlang.org/play?ts=3.7.5&q=247#code/GYVwdgxgLglg9mABMMAKAlIg3ogTgUyhFyQHIBDAIwlIG5EBfAWAChWZhFUVMtmWOXAIQ9s-QdyQBeKYii4Q+XuM6pUIsJhlyFSsa1ZA

💻 Code

function fn() { return 'abc'; }

if (fn) {}  // ERROR: This condition will always ...
if (!fn) {} // not reported
if (fn == true) {} // ERROR: This condition will always ...
if ((!fn) == true) {} // ERROR: This condition will always ...

🙁 Actual behavior

No warning

🙂 Expected behavior

“This condition will always …” warning

[concavelenz]

Detecting fn when the author intended fn() is valuable. TSC already catches “if (fn) …” is great, I’d hope that catching "if (!fn) " would also be seen as worthwhile as !x is a common use of a result value in a boolean expression.

[DanielRosenwasser]

So this is definitely confusing, especially because the error messages are similar; but my understanding is

• if (fn) is getting an error for always-defined functions. That’s expected.
• if (!fn == true) is getting an error for something called comparability which checks for overlap between two types. The check is expected , but I think the result of !fn isn’t.

I think we’re trying to be “too smart” in the second example though. I would expect !fn would just become boolean (not false), and would pass the check.