#4 - fix: protocol activity statistics stay zero — ponder watches wrong contract addresses - johba/harb

johba commented

2026-04-05 17:14:12 +00:00

Owner

Problem

On the app page, protocol activity statistics stay zero on all parameters (except holder count) even after trades and staking.

Reported in codeberg#1193.

Root cause (three issues)

1. Missing `LM_ADDRESS` in ponder's `.env.local`

containers/bootstrap.sh write_ponder_env() writes KRAIKEN_ADDRESS and STAKE_ADDRESS but does not write LM_ADDRESS. Without it, ponder falls back to a hardcoded default in ponder.config.ts that has no code deployed on this Anvil fork. The actual LiquidityManager is at a different address.

Result: 5 LiquidityManager events on chain (Recentered, EthScarcity/EthAbundance) are completely missed. This causes: lastRecenterTimestamp=0, recentersLastDay=0, lastEthReserve=0.

2. Missing `POOL_ADDRESS` in ponder's environment

POOL_ADDRESS is hardcoded in services/ponder/src/kraiken.ts and stake.ts as a Base Sepolia address. On local Anvil, the pool is at a different address. This means:

Swap detection (isBuy = from.toLowerCase() === POOL_ADDRESS) never matches
No transactions recorded
updateEthReserve() reads WETH balance of a non-existent pool (returns 0)

3. Ring buffer never populated for early events

The initial Transfer events happen within 13 blocks of deployment. checkBlockHistorySufficient() requires 100 blocks since deployment before populating the ring buffer. All initial events hit the early-return path that updates totalMinted/totalBurned but skips the ring buffer, so rolling window stats (mintedLastDay, mintedLastWeek, etc.) compute to zero.

Why holder count works

holderCount is tracked directly in the Kraiken:Transfer handler — it increments/decrements when balances change from/to zero. It does NOT depend on the ring buffer, pool address, or LM address.

Fix

containers/bootstrap.sh write_ponder_env() must include LM_ADDRESS and POOL_ADDRESS from the deployed contract addresses
services/ponder/src/kraiken.ts and stake.ts should read POOL_ADDRESS from env instead of hardcoding it
Consider reducing the 100-block history check for local dev environments, or seeding the ring buffer on first event regardless

Reproduction

curl -sf http://localhost:42069/graphql -H 'Content-Type: application/json' \
  -d '{"query": "{ stats(id: \"0x01\") { holderCount mintedLastDay burnedLastDay recentersLastDay lastEthReserve } }"}' | jq .
# holderCount=2, everything else=0

Files

containers/bootstrap.sh — write_ponder_env() missing LM_ADDRESS and POOL_ADDRESS
services/ponder/src/kraiken.ts — hardcoded POOL_ADDRESS
services/ponder/src/stake.ts — hardcoded POOL_ADDRESS
services/ponder/ponder.config.ts — LM_ADDRESS fallback default
services/ponder/src/helpers/ringBuffer.ts — checkBlockHistorySufficient() 100-block threshold

## Problem On the app page, protocol activity statistics stay zero on all parameters (except holder count) even after trades and staking. Reported in [codeberg#1193](https://codeberg.org/johba/harb/issues/1193). ## Root cause (three issues) ### 1. Missing `LM_ADDRESS` in ponder's `.env.local` `containers/bootstrap.sh` `write_ponder_env()` writes `KRAIKEN_ADDRESS` and `STAKE_ADDRESS` but **does not write `LM_ADDRESS`**. Without it, ponder falls back to a hardcoded default in `ponder.config.ts` that has no code deployed on this Anvil fork. The actual LiquidityManager is at a different address. Result: 5 LiquidityManager events on chain (Recentered, EthScarcity/EthAbundance) are completely missed. This causes: `lastRecenterTimestamp=0`, `recentersLastDay=0`, `lastEthReserve=0`. ### 2. Missing `POOL_ADDRESS` in ponder's environment `POOL_ADDRESS` is hardcoded in `services/ponder/src/kraiken.ts` and `stake.ts` as a Base Sepolia address. On local Anvil, the pool is at a different address. This means: - Swap detection (`isBuy = from.toLowerCase() === POOL_ADDRESS`) never matches - No transactions recorded - `updateEthReserve()` reads WETH balance of a non-existent pool (returns 0) ### 3. Ring buffer never populated for early events The initial Transfer events happen within 13 blocks of deployment. `checkBlockHistorySufficient()` requires 100 blocks since deployment before populating the ring buffer. All initial events hit the early-return path that updates `totalMinted`/`totalBurned` but skips the ring buffer, so rolling window stats (`mintedLastDay`, `mintedLastWeek`, etc.) compute to zero. ## Why holder count works `holderCount` is tracked directly in the `Kraiken:Transfer` handler — it increments/decrements when balances change from/to zero. It does NOT depend on the ring buffer, pool address, or LM address. ## Fix 1. `containers/bootstrap.sh` `write_ponder_env()` must include `LM_ADDRESS` and `POOL_ADDRESS` from the deployed contract addresses 2. `services/ponder/src/kraiken.ts` and `stake.ts` should read `POOL_ADDRESS` from env instead of hardcoding it 3. Consider reducing the 100-block history check for local dev environments, or seeding the ring buffer on first event regardless ## Reproduction ```bash curl -sf http://localhost:42069/graphql -H 'Content-Type: application/json' \ -d '{"query": "{ stats(id: \"0x01\") { holderCount mintedLastDay burnedLastDay recentersLastDay lastEthReserve } }"}' | jq . # holderCount=2, everything else=0 ``` ## Files - `containers/bootstrap.sh` — `write_ponder_env()` missing LM_ADDRESS and POOL_ADDRESS - `services/ponder/src/kraiken.ts` — hardcoded POOL_ADDRESS - `services/ponder/src/stake.ts` — hardcoded POOL_ADDRESS - `services/ponder/ponder.config.ts` — LM_ADDRESS fallback default - `services/ponder/src/helpers/ringBuffer.ts` — `checkBlockHistorySufficient()` 100-block threshold