fix: slashable window boundaries

[8sunyuan]

Fixing slashable windows here by ensuring that from
[block.number, block.number + MIN_WITHDRAWAL_DELAY_BLOCKS] and
[block.number, block.number + DEALLOCATION_DELAY] are both the same window.
Since we want these two values MIN_WITHDRAWAL_DELAY_BLOCKS and DEALLOCATION_DELAY to be configured to be the same in deployments, we make the following changes:

1. Upon withdrawal completion, we now have a slashableUntil block which is withdrawal.startBlock + MIN_WITHDRAWAL_DELAY_BLOCKS which is the last block the staker's withdrawal is slashable for. That means the first block they can complete withdraw is the block following hence the strict inequality

uint32 slashableUntil = withdrawal.startBlock + MIN_WITHDRAWAL_DELAY_BLOCKS;
require(slashableUntil < uint32(block.number), WithdrawalDelayNotElapsed());

2. For deallocations, we have effectBlock set as block.number + DEALLOCATION_DELAY currently but that is the block that the deallocation is no longer slashable and can be completed. We fix this to ensure the slashable window of [block.number, block.number + DEALLOCATION_DELAY] by making the effectBlock = block.number + DEALLOCATION_DELAY + 1.

3. For DelegationManager._getSlashedSharesInQueue we want also ensure a slashable window of [block.number - MIN_WITHDRAWAL_DELAY_BLOCKS, block.number] including blocknumber block.number - MIN_WITHDRAWAL_DELAY_BLOCKS. So since a subtraction is done of curCumulativeScaledShares - prevCumulativeScaledShares, we make the snapshot.lookup of prevCumulativeScaledShares to be -1 to include that block itself.

[gpsanant]

this should be >=

alternatively, making it uint32(block.number) > slashableUntil and switching the inners of the conditionals is cleaner

[8sunyuan]

_getSlashingFactors gets the current slashing factor, so the logic here is that if slashableUntil is at some future block, then we simply read the current slashing factors. Else we read it at the exact slashableUntil block

[ypatil12]

Its clearer if we go >=, but the underlying logic is still the same

[8sunyuan]

I get now what you guys are saying now and agreed.
Updated now e6c66cd

[ypatil12]

Note: every comparison to effectBlock is < effectBlock. There is some minor difference in naming (and resulting comparisons) because we do:

• < effectBlock
• <= slashableUntil

Nomenclature does make sense here though

[ypatil12]

test_Revert_WhenWithdrawalDelayNotPassed validates this

[ypatil12]

6ffb6c3

[ypatil12]

Burning test: 8d380fa