gdb set value at address of a float

I am learning about floating point number representation and I have the following C++ code:

#include <float.h>
int main()
{
    unsigned int a = 8;
    float f = 1;
    float fmax = FLT_MAX;
    
    float n = 3.402823669e+38;

    return 0;
}

When I print the value of f as binary in gdb like x/tw &f I get:
0 01111111 00000000000000000000000 (1 sign bit, 8 exp bits, 23 mantissa bits).
In hexadecimal it is 0x3f800000.

Now I want to see which number I will get if I make the exponent as large as possible.

So I do set *(&f) = 0x7f800000.

However, instead of getting the expected
0 11111111 00000000000000000000000,

I get something completely arbitrary like
0 10011101 11111110000000000000000,
which is 0x4eff0000, which is not the value I set.

This happens also if the set result doesn't exceed FLT_MAX.

Changing unsigned int a in this way works fine.

What could be a possible explanation to this behavior?

> What could be a possible explanation to this behavior?

OP assigned the value of 2139095040.0f to f.

*(&f) = 0x7f800000 is like *(&f) = 2139095040.0f.

2139095040.0f is encoded in a float with hex pattern 0x4eff0000.

No surprise.