Function redefinition doesn't always work

[carlobaldassi]

I found that sometimes function overwriting does not work right away. The method redefinition warning gets triggered, but the old function definition is used.
Here is a minimal test case which shows the bug (the last call fails, but calling f() again then produces 2 as expected):

f() = 1

# at least two calls are required to trigger the bug
f()
f()

f() = 2

@assert f() == 2 # fails

This is on version 0.5.0-dev+4882.

[vtjnash]

I was wondering when someone would notice this. (not a duplicate of #265)

[StefanKarpinski]

You want to explain what you changed, @vtjnash?

[oxinabox]

I had been finding this also, kinda thought either it was by design (no redefining consts) or that someone else had reported it.

[tkelman]

You want to explain what you changed, @vtjnash?

Apparently not?

[vtjnash]

What better way to explain?

[tkelman]

A function with no comments or tests doesn't actually explain what was going wrong in a way that anyone other than you understands.