Block number inconsistency between bench tests and runs

[ggwpez]

In benchmark-tests and runs the block number is set implicitly to 1 after the setup if it was 0.
This is confusing since normally the user must set it manually in the TestExternalties.
Could be changed to either:

• Never touch the block-number and let the user deal with it - same as in tests
• Always set the block-number to one before the setup

Two benchmark tests to demonstrate the current behaviour:

block_jumps_to_one {
        // Starts at block 0.
	assert_eq!(frame_system::Pallet::<T>::block_number(), 0u32.into());
}: {
        // Why does it jump to one?
	assert_eq!(frame_system::Pallet::<T>::block_number(), 1u32.into());
}

block_does_not_jump_back {
        // Jump to block 2.
	frame_system::Pallet::<T>::set_block_number(2u32.into());
}: {
        // Why does it NOT jump back to one?
	assert_eq!(frame_system::Pallet::<T>::block_number(), 2u32.into());
}

Came up again here https://substrate.stackexchange.com/questions/5516

[bragov4ik]

+1, I was surprised when exactly the same code behave differently in benchmarks and tests 😢