How the fuck do you accurately estimate development time , testing time , fixes time ?

Know the system,
Know the requirement,
Know the changes required.

From here you can estimate pretty accurately.

If you don't know all 3, your under estimating.

@C0D4 that and this estimate multiplied by 2.5?

@heyheni π is the popular choice.

Guess x2

@dan-pud plus the managerial "no way! this has to be done in 2x instead of 3x"?

4

XP and some randomness

It is never accurate hence the name.

I get that managers take estimates as deadlines, but that's because they are illiterate, don't be hard on them!

Yes

Like others said: you cannot accurately estimate it, better have your estimates too big than too small of course. But you can make use of "backpropagation" to adjust estimations while you're working on the tasks. And also separating one large task into smaller tasks can make it more predictable, or at the very least more flexible to adjust your estimations while you're working on them.
You could also use something like Scrum to see how your estimations hold up to reality and it could warn you that you thought too lightly on a certain task for example.

Experience. Your experience. Just because the team thinks it should be a 2, throw your 3 anyway if that's the way you feel

Break everything down to the smallest operation/component you can think of, estimate each of them with Fibonacci numbers.

@ReverendLovejoy I've always done Fibonacci, but this new group I'm with doesn't. You get wild 4's from time to time. Really throws me off

@coder-guy can you prove them wrong by comparing actual work done vs estimated work ?