Follow-up from #202 (and unaddressed by stacked #285, #289).
What
surfpool_jupiter_swap_roundtrip (now run_idl_roundtrip after #285's refactor) starts a surfpool mainnet-fork instance, assigns it to _manager, then runs the local visual-sign parser on a synthetic transaction whose data is [discriminator] + [0u8; 32]. The assertion is just !payload.fields.is_empty(). The surfpool instance is dropped without ever being queried, simulated against, or otherwise used.
Where
src/chain_parsers/visualsign-solana/tests/common/mod.rs (post-#289) — search for let _manager = SurfpoolManager::start(...).
Why it matters
The PR titled "surfpool mainnet-fork integration for built-in IDL fuzz path" advertises mainnet-fork integration, but as written the test is an IDL-discriminator fuzz that happens to start surfpool as a slow side-effect. CI minutes and Helius RPC credits are spent starting a fork that contributes nothing to the test's signal. A real integration would, e.g.:
- Use surfpool's
simulateTransaction against a real on-chain program ID (loaded from the fork).
- Pull an actual recent on-chain transaction via the forked RPC and feed it through the parser.
- Assert specific output shape (named accounts resolved against real mint addresses, etc.).
Alternative: rename to reflect what it actually does (e.g. idl_fuzz not surfpool_fuzz) and drop the surfpool startup from this path; keep surfpool-backed tests as a separate, narrower suite.
Repro
cargo test -p visualsign-solana --test surfpool_fuzz surfpool_idl_jupiter -- --ignored --nocapture
The surfpool process starts, the parser runs in-process on synthetic bytes, the test passes regardless of what the fork is doing.
Follow-up from #202 (and unaddressed by stacked #285, #289).
What
surfpool_jupiter_swap_roundtrip(nowrun_idl_roundtripafter #285's refactor) starts a surfpool mainnet-fork instance, assigns it to_manager, then runs the local visual-sign parser on a synthetic transaction whose data is[discriminator]+[0u8; 32]. The assertion is just!payload.fields.is_empty(). The surfpool instance is dropped without ever being queried, simulated against, or otherwise used.Where
src/chain_parsers/visualsign-solana/tests/common/mod.rs(post-#289) — search forlet _manager = SurfpoolManager::start(...).Why it matters
The PR titled "surfpool mainnet-fork integration for built-in IDL fuzz path" advertises mainnet-fork integration, but as written the test is an IDL-discriminator fuzz that happens to start surfpool as a slow side-effect. CI minutes and Helius RPC credits are spent starting a fork that contributes nothing to the test's signal. A real integration would, e.g.:
simulateTransactionagainst a real on-chain program ID (loaded from the fork).Alternative: rename to reflect what it actually does (e.g.
idl_fuzznotsurfpool_fuzz) and drop the surfpool startup from this path; keep surfpool-backed tests as a separate, narrower suite.Repro
cargo test -p visualsign-solana --test surfpool_fuzz surfpool_idl_jupiter -- --ignored --nocaptureThe surfpool process starts, the parser runs in-process on synthetic bytes, the test passes regardless of what the fork is doing.