What is the correct way to print a float?

[Kempelen]

I have a program that printf a float that is returned by a function like this:

printf("%3.1f\n", myfloat());
...
float myfloat(void) {
    code....
}

In w32 my program has been running ok, but when run in a w64, it crash at the printf. If I change the definition of the function to return double it runs ok.

For my surprise, in printf format string - Wikipedia, the free encyclopedia , in the "type" section I see there is no definition to print a float, but %f is for double in fixed point.

So the question is what is the correct way to printf a float? must I cast? or is it better to return double? if so, does it have a performance penalty?

After seeing the wikipedia I am a bit confused about it.....

[c99tutorial]

Try calling printf like this

printf("%3.1f\n", (double)myfloat());

Conversion only happens only when printing. Even if you use exclusively double throughout your program rather than float, this is often not a performance difference. However, the only valid way to know is to profile your code execution using float vs double. But such minor details are normally not speed critical compared to the overall approach or algorithm being used.

[nonoob]

Did you forward-declare the function? Otherwise you may have been lucky that the compiler's 'int' guess happens to have the same size as 'float' data type in a particular platform.

[Barney McGrew]

It shouldn't matter. If you pass in a float it will be automatically converted to a double. Chances are your implementation's broken.

[Adak]

Maybe it's just your example, but your myfloat() function doesn't show an actual return value.