gas2500x test plan

Field test for the Gas2500x tier (625 gwei) shipped by three XDPoSChain
commits on dev-upgrade (baseline 5501a1f1e5, all three included):

• 7823fc7f (#2516) — Gas2500xBlock fork: the pool floor, the chain default gas price and the EIP-1559 base fee move 12.5 gwei → 625 gwei, and --gasprice can no longer move the floor.
• 68695806 (#2532) — on a floor rise the pool sweeps every non-special tx below the new floor from pending and queue; swept txs never come back.
• 06582401 (#2541) — the local tracker holds back below-floor local txs and revives them when the floor drops; below-floor rejects never enter the tracker.

A test result: https://github.com/gzliudan/Local_DPoS_Setup/blob/gas2500x/results/gas2500x-20260905-182603.log

Test environment

• OS: Ubuntu 24.04.4 LTS (Linux 7.0.0, x86_64)
• bash: GNU bash 5.2.21
• jq: jq 1.7
• cast: foundry cast 1.8.1
• curl: curl 8.5.0
• Go: go 1.25.14 (builds the XDC binaries)

Quick start

cd ~/XDPoSChain
git use dev-upgrade
make all                            # once: XDC binaries

cd ~
git clone https://github.com/gzliudan/Local_DPoS_Setup/
cd Local_DPoS_Setup/
git use gas2500x
cp .env.sample .env                 # once: sender/node keys (S1–S5, P3)
# Fill the real keys in `.env`
./gas2500x-run.sh                   # reset → start network → 53 cases → stop
./stop-network.sh                   # only after an interrupted run

Fill the real keys in .env — the three masternode keys
(PRIVATE_KEY_0/1/2) are bound to genesis.json: their addresses are the
masternodes baked into the genesis extraData (and funded in alloc),
so they must stay exactly as shipped; generating fresh keys here yields
a node that can never sign. PRIVATE_KEY_3 (observer pn3) and
TXGEN_KEY_1..5 (senders S1–S5) may be any funded-able raw hex keys.

The runner does the whole lifecycle itself — wipes the datadirs, starts the
network, runs every case, stops the network.

Topology

Masternodes keep sealing the whole run — 2/3 quorum is intact. All pool
cases submit via pn3, so tracker/journal state lands there and the rewind
experiment touches pn3 only.

Key numbers

• Gas50x floor: 12.5 gwei
• Gas2500x floor: 625 gwei
• gas2500xBlock: 120
• Tracker recheck: 10s after start, then every 60 s (locals.recheckInterval)

Test scripts

Local_DPoS_Setup/
├── gas2500x-run.sh          # the only entry point
├── gas2500x-test.md         # this document
├── results/
│   └── gas2500x-<ts>.log    # runner transcript (start/end tally inside)
└── tests/
    ├── gas2500x-lib.sh      # shared helpers
    ├── t01.sh … t23.sh      # pre-fork
    ├── t24.sh … t34.sh      # fork sweep
    ├── t35.sh … t42.sh      # post-fork probes
    ├── t43.sh … t45.sh      # rewind saga
    └── t46.sh … t53.sh      # creation matrix

Test cases

53 cases, one verifiable result each. The fork (block 120) splits them into
mirrored pairs, one case per side (T02/T27 reject, T09/T29 at-floor seal,
T13/T32 replace, T21 parks / T30 seals the survivor, T15/T42 default price);
observation-only cases add no submissions (T10, T22, T24–T26, T34, T44,
T45); the survivor chain T21→T24→T30 and the rewind saga T43→T44→T45 are
fixed state chains (via /tmp markers); the creation matrix t46–t53 cases
are mutually independent. Missed the pre-fork window? Just re-run the runner
— it resets the chain.

One benign interference: every ~30 s each masternode broadcasts a consensus
signing tx (gasPrice 0) that transiently enters every pool as
executable-special for ~2 s. The pending-side assertions (T11, T12, T34,
T45) therefore count only non-signing txs (shape-based filter).

Out of scope (unit-test covered): TRC21/XDCx tier pricing, concurrent
TrackAll races, Osaka gas-cap discard, Gas2500xBlock == nil, special txs.

Case names are verb-object and paren-free. The table below is generated
from the scripts' begin_case args — regenerate it after renaming a case.