enabled and fixed -Wdouble-promotion Clang warnings

[firewave]

No description provided.

[danmar]

Clang/GCC doesn't implement their checkers properly...

Code 1:

void dostuff(double x);

void f() {
    dostuff(0.1f);
}

Clang says:

test1.c:5:13: warning: implicit conversion increases floating-point precision: 'float' to 'double' [-Wdouble-promotion]

Code 2:

void dostuff(double x);

void f() {
    dostuff(static_cast<double>(0.1f));
}

.. after some time you have this ..

Code 3:

void dostuff(float x);

void f() {
    dostuff(static_cast<double>(0.1f));
}

And the compilers are silent about this. By turning on -Wdouble-promotion this is what you get. There was no problem in Code 1 but in Code 3 there is a performance problem. And it hurts the readability.

[firewave]

Clang/GCC doesn't implement their checkers properly...

In that case please file an upstream report.

Looks like this might also be related to llvm/llvm-project#93288.

[danmar]

In that case please file an upstream report.

Nah, I doubt they would care about it.

[danmar]

I reject this because it can lead to performance and readability issues in the long run.

[danmar]

I don't think this "fix" the issue.. there is still the promotion that clang warned about. This code will work technically exactly the same as before as far as I see.

[firewave]

Nah, I doubt they would care about it.

Great attitude. Imagine people had the attitude towards false positives in Cppcheck.

Well, you might not imagine. They have that 100%. They just suppress the FP and if at some point there are too many suppressions or it gets too slow they just completely drop it completely because it "offers no value". That's how any tooling (not just open source) starts to deteriorate (even in cases where the devs actually care about the application and bug reports).

Well, they basically were too lazy to file a report and found me as the idiot which did their work for them because I cared.

That is actually the scenario because I got involved again because I wanted it to improve, so other people see that it actually has value and should be kept. See where that got me ...

And there has also been cases where people (including me) made something out to be a "false positive" but it turned out it was not and at the end you actually learned something new....

[danmar]

Well, you might not imagine. They have that 100%.

Yes I have seen it myself also.