TMin - whether TMin overflows

At first consider the principle of two's complement subtraction: $a{-}^{t}b=a{+}^t$, and the negative of TMin is still TMin, so I think TMin - TMin should be equivalent to TMin+TMin, so it is naturally an overflow.
But when learning the setl instruction, the book says that setb is judged according to SF^OF, so I think, cmpl $0x80000000, $0x80000000 (ie operation $0x80000000 - $0x80000000), OF=1, SF=0, then the result of setb should beis 1, so you get the paradox $0x80000000 < $0x80000000.
I struggled for a long time until I actually typed the assembly code:

get_OFxorSF:movl %ecx, %eaxsubl %edx, %eaxsetl %almovzbq %al, %eaxretget_SF:movl %ecx, %eaxsubl %edx, %eaxsets %almovzbq %al, %eaxret

It turns out that when 0x80000000-0x80000000, both OF^SF and SF are set to 0.Therefore OF is also 0, ie TMin-TMin does not overflow.
And when I change the subl of the above command to addl, the result is: OF^SF is set to 1, SF is set to 0, that is, >TMin+TMin overflowed.
In conclusion, this kind of entanglement is due to insufficient understanding of the definition of overflow. Overflow means that the result obtained by computer calculation is different from the result obtained by simple arithmetic operation.So of course TMin-TMin=0 will not generate overflow.