Parent epic: https://github.com/DIG-Network/dig_ecosystem/issues/3166
Sibling: #412 (the lifecycle family this came out of)
Task
bondable_pairs counts every served row except BondState::Withheld. Establish whether
BondState::Disabled belongs in that exclusion too, and make the answer explicit in code and in the
doc — either by excluding it, or by stating why a Disabled row still locks collateral.
Context
Found by the PR #419 gate round, out of that re-gate's scope, so it was left rather than guessed at.
bondable_pairs — crates/dig-node-service/src/control.rs, the count that feeds
control.collateral.buffer.BondState::Disabled — crates/dig-node-service/src/mirror/pass.rs:104.- The function's own doc asserts "every state except
Withheld", so this is a contract
question rather than an obvious bug. Nothing currently in the tree establishes that Disabled
is bondable — it is asserted by the doc and by nothing else.
Disabled describes a capsule this node deliberately does not advertise and whose coin is
being reclaimed — which is, on its face, the same situation as Withheld: nothing to lock
collateral against.
Why this is a ticket and not a comment
It fails in the money-inflating direction, and it is the same direction as the defect the
Withheld exclusion was added to fix. If a Disabled row locks nothing, the buffer advice
over-states what the operator must hold — telling someone to reserve $DIG against a bond that is
on its way out. A collateral figure that is wrong upward is a surface making a claim about the
operator's money.
Scope
Answer the contract question first, then encode it. If Disabled is excluded, the test at
control.rs:3773 extends with a fifth row; if it is not excluded, the doc gains the sentence
saying why, so the next reader does not re-open this.
Do not widen it into a general audit of the state machine.
Evidence
The existing test is load-bearing and shows the shape to copy — reverting the filter to
observation.states.len() is red at left: 4 / right: 3. Whichever way this resolves, it needs a
test that distinguishes the contract from "count everything" and from "count the last row" — a
two-row fixture answers 1 under several wrong predicates, which is why the current test uses four
rows carrying three distinct states.
Parent epic: https://github.com/DIG-Network/dig_ecosystem/issues/3166
Sibling: #412 (the lifecycle family this came out of)
Task
bondable_pairscounts every served row exceptBondState::Withheld. Establish whetherBondState::Disabledbelongs in that exclusion too, and make the answer explicit in code and in thedoc — either by excluding it, or by stating why a
Disabledrow still locks collateral.Context
Found by the PR #419 gate round, out of that re-gate's scope, so it was left rather than guessed at.
bondable_pairs—crates/dig-node-service/src/control.rs, the count that feedscontrol.collateral.buffer.BondState::Disabled—crates/dig-node-service/src/mirror/pass.rs:104.Withheld", so this is a contractquestion rather than an obvious bug. Nothing currently in the tree establishes that
Disabledis bondable — it is asserted by the doc and by nothing else.
Disableddescribes a capsule this node deliberately does not advertise and whose coin isbeing reclaimed — which is, on its face, the same situation as
Withheld: nothing to lockcollateral against.
Why this is a ticket and not a comment
It fails in the money-inflating direction, and it is the same direction as the defect the
Withheldexclusion was added to fix. If aDisabledrow locks nothing, the buffer adviceover-states what the operator must hold — telling someone to reserve $DIG against a bond that is
on its way out. A collateral figure that is wrong upward is a surface making a claim about the
operator's money.
Scope
Answer the contract question first, then encode it. If
Disabledis excluded, the test atcontrol.rs:3773extends with a fifth row; if it is not excluded, the doc gains the sentencesaying why, so the next reader does not re-open this.
Do not widen it into a general audit of the state machine.
Evidence
The existing test is load-bearing and shows the shape to copy — reverting the filter to
observation.states.len()is red atleft: 4 / right: 3. Whichever way this resolves, it needs atest that distinguishes the contract from "count everything" and from "count the last row" — a
two-row fixture answers
1under several wrong predicates, which is why the current test uses fourrows carrying three distinct states.