Further examine necessity of clipping on withdrawals

[cwgoes]

Ref https://github.com/cosmos/cosmos-sdk/pull/3033/files#r239978938

[cwgoes]

Possibly this is addressed by #2958.

[rigelrozanski]

Sure we ought to specify this better in new docs. In the Dec=>Int PR we always ensure that the validator will be emptied aka we do not burn tokens ever.

[cwgoes]

My concern is as to why this was happening in the first place - seems like some delegator must have previously received more tokens than they ought to have (which is very concerning).

[rigelrozanski]

What's happening in Jae's PR is not completely obvious to me, however in the Dec=>Int PR there is a clear situation which gives rise to a delegator share being "worth" an uneven amount of tokens THUS when they withdraw, the remaining tokens should be left in the validator, making all the other delegator shares now worth slightly more

[jackzampolin]

Is this closed by F1?

[cwgoes]

Is this closed by F1?

No, this has nothing to do with fee distribution, the "withdrawal" referred to is withdrawal of tokens from a validator (in the staking module).

[rigelrozanski]

Do we even clip any more? - during the Dec=>Int this has been consolidated to 1 or 2 straight forward places

[jackzampolin]

@rigelrozanski thats a great question!

[rigelrozanski]

Just investigated this... within staking clipping is isolated to the (validator) RemoveDelShares
function

cosmos-sdk/x/staking/types/validator.go

Lines 370 to 384 in b5e245f

	if remainingShares.IsZero() {
	// last delegation share gets any trimmings
	issuedTokens = v.Tokens
	v.Tokens = sdk.ZeroInt()
	} else {
	// leave excess tokens in the validator
	// however fully use all the delegator shares
	issuedTokens = v.DelegatorShareExRate().Mul(delShares).TruncateInt()
	v.Tokens = v.Tokens.Sub(issuedTokens)
	if v.Tokens.IsNegative() {
	panic("attempting to remove more tokens than available in validator")
	}
	}

I think it's pretty straightforward why this needs to happen, under the situation that slashing caused delegator's shares to be worth an uneven amount of tokens. Again we are rounding at the nano-atom level.

I'm confident that this is correct and no reason for concern.