Skip to content

fix(design): cold-start emit current epoch #119

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Merged
echobt merged 1 commit into from
Aug 10, 2026
Merged

fix(design): cold-start emit current epoch #119

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
1 commit
Select commit
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
43 changes: 40 additions & 3 deletions crates/design-challenge-task/src/emit.rs
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -16,10 +16,16 @@ pub struct DesignEmitPlan {
pub pin_block: u64,
}

/// Max epochs to walk back on catch-up (Finney public RPC prunes ~256 blocks).
const MAX_CATCHUP_EPOCHS: u64 = 16;

/// Decide whether/which epoch the design filler should emit.
///
/// - Catch up `last_emitted+1` when behind by more than one epoch (repairs the
/// end-of-epoch relabel race that skipped alternate epochs in prod).
/// - **Cold start** (`last_emitted == 0`): emit the **current** epoch immediately.
/// The in-process cursor resets to 0 on every process boot; walking from
/// epoch 1 pins a pruned block and fails with `SubnetOwnerHotkey not found`.
/// - Catch up `last_emitted+1` when behind (capped to [`MAX_CATCHUP_EPOCHS`])
/// so end-of-epoch relabel skips can recover without exceeding prune depth.
/// - Otherwise wait until the last [`DESIGN_EMIT_LATE_BLOCKS`] of the current
/// epoch so admin awards can submit Score leaves first.
#[must_use]
Expand All @@ -34,12 +40,24 @@ pub fn design_emit_plan(
return None;
}
let tempo = tempo.max(1);
// Cold start after deploy/restart: do not catch up from epoch 1.
if last_emitted == 0 {
return Some(DesignEmitPlan {
epoch: current_epoch,
pin_block: current_last_epoch_block,
});
}
if last_emitted >= current_epoch {
return None;
}
// Sequential catch-up for skipped epochs (award path / boundary race).
if last_emitted + 1 < current_epoch {
let target = last_emitted + 1;
let gap = current_epoch.saturating_sub(last_emitted);
let target = if gap > MAX_CATCHUP_EPOCHS {
current_epoch.saturating_sub(MAX_CATCHUP_EPOCHS)
} else {
last_emitted + 1
};
let epochs_back = current_epoch.saturating_sub(target);
let pin_block = current_last_epoch_block.saturating_sub(epochs_back.saturating_mul(tempo));
Comment on lines 54 to 62

@coderabbitai coderabbitai Bot Aug 10, 2026

Copy link
Copy Markdown

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

🩺 Stability & Availability | 🟠 Major | ⚡ Quick win

Calculate the catch-up limit from block retention.

MAX_CATCHUP_EPOCHS does not bound the age of pin_block. With the new test value tempo = 360, current_epoch - 16 produces a pin block 5,760 blocks old. This exceeds the approximately 256-block retention stated at line 19, so the RPC can still reject the planned emission.

Cap epochs_back by the supported pin-block age divided by tempo. If one epoch exceeds that age, do not select a historical epoch. Update the cap test to enforce the block-age invariant.

🤖 Prompt for AI Agents 
Verify each finding against current code. Fix only still-valid issues, skip the
rest with a brief reason, keep changes minimal, and validate.

In `@crates/design-challenge-task/src/emit.rs` around lines 54 - 62, Update the
catch-up calculation in the emission logic around MAX_CATCHUP_EPOCHS so
epochs_back is also bounded by the supported pin-block retention divided by
tempo. When tempo exceeds the retention age, avoid selecting a historical epoch;
otherwise ensure
current_last_epoch_block.saturating_sub(epochs_back.saturating_mul(tempo)) stays
within the retention limit, and revise the cap test to assert this block-age
invariant.

return Some(DesignEmitPlan {
Expand Down Expand Up @@ -82,6 +100,25 @@ mod tests {
assert_eq!(p.pin_block, 8_815_687 - (24423 - 24413) * 360);
}

#[test]
fn emit_plan_cold_start_emits_current_immediately() {
let p = design_emit_plan(0, 24424, 10, 360, 8_816_047).unwrap();
assert_eq!(
p,
DesignEmitPlan {
epoch: 24424,
pin_block: 8_816_047
}
);
}

#[test]
fn emit_plan_caps_catchup_to_prune_window() {
let p = design_emit_plan(100, 24424, 10, 360, 8_816_047).unwrap();
assert_eq!(p.epoch, 24424 - 16);
assert_eq!(p.pin_block, 8_816_047 - 16 * 360);
}
Comment on lines +103 to +120

@coderabbitai coderabbitai Bot Aug 10, 2026

Copy link
Copy Markdown

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

🗄️ Data Integrity & Integration | 🟠 Major | 🏗️ Heavy lift

🧩 Analysis chain

🏁 Script executed:

#!/bin/bash
set -eu

printf '%s\n' '--- file inventory ---'
git ls-files 'crates/design-challenge-task/src/emit.rs' 'crates/design-challenge/src/lib.rs' 'crates/design-challenge*' | sed -n '1,160p'

printf '%s\n' '--- outlines ---'
ast-grep outline crates/design-challenge-task/src/emit.rs --lang rust 2>/dev/null | sed -n '1,220p'
ast-grep outline crates/design-challenge/src/lib.rs --lang rust 2>/dev/null | sed -n '1,220p'

printf '%s\n' '--- relevant symbols and assertions ---'
rg -n -C 3 \
  'design_emit_plan|DesignEmitPlan|emit_plan|sealed|seal|raw weight|raw_weight|leaf|failure probe|challenge-specific|intake|submit' \
  crates/design-challenge-task/src/emit.rs \
  crates/design-challenge/src/lib.rs \
  crates/design-challenge-task \
  crates/design-challenge \
  2>/dev/null | sed -n '1,320p'

Repository: BaseIntelligence/base

Length of output: 24732

🏁 Script executed:

#!/bin/bash
set -eu

for f in \
  crates/design-challenge/tests/e2e_sim.rs \
  crates/design-challenge/src/orchestrator.rs \
  crates/design-challenge/src/host_sim.rs \
  crates/design-challenge/src/lib.rs
do
  if [ -f "$f" ]; then
    printf '\n--- %s (%s lines) ---\n' "$f" "$(wc -l < "$f")"
    ast-grep outline "$f" --lang rust 2>/dev/null | sed -n '1,260p'
  fi
done

printf '\n--- test names and flow terms ---\n'
rg -n -C 4 \
  '#\[test\]|#\[tokio::test\]|failure|probe|intake|validate|leaf|weight|seal|sealed|emit_signed_leaf_set|submit_signed_leaf_set|GatewayClient' \
  crates/design-challenge/tests crates/design-challenge/src crates/challenge-common \
  2>/dev/null | sed -n '1,520p'

printf '\n--- shared helper locations ---\n'
rg -n -C 3 \
  'pub (async )?fn (emit_signed_leaf_set|submit_signed_leaf_set)|struct GatewayClient|fn seal|sealed' \
  crates 2>/dev/null | sed -n '1,360p'

Repository: BaseIntelligence/base

Length of output: 50377

🏁 Script executed:

#!/bin/bash
set -eu

printf '%s\n' '--- e2e_sim.rs ---'
cat -n crates/design-challenge/tests/e2e_sim.rs

printf '%s\n' '--- orchestrator emitter and execution ranges ---'
sed -n '500,760p' crates/design-challenge/src/orchestrator.rs | nl -ba
sed -n '760,960p' crates/design-challenge/src/orchestrator.rs | nl -ba
sed -n '960,1110p' crates/design-challenge/src/orchestrator.rs | nl -ba

printf '%s\n' '--- shared submit implementation and tests ---'
cat -n crates/challenge-common/src/submit.rs | sed -n '1,360p'
printf '%s\n' '--- leaf emit implementation and tests ---'
cat -n crates/challenge-common/src/leaf_emit.rs | sed -n '1,260p'

printf '%s\n' '--- gateway sealer routes and tests ---'
rg -n -C 8 \
  'Router|route|seal|sealed|raw|weight|test|#\[tokio::test\]|#\[test\]' \
  crates/gateway/src/sealer.rs crates/gateway/src/api.rs crates/gateway/tests \
  2>/dev/null | sed -n '1,520p'

printf '%s\n' '--- challenge flow markers in tests ---'
python3 - <<'PY'
from pathlib import Path
files = list(Path("crates/design-challenge").rglob("*.rs"))
terms = ("failure", "probe", "intake", "validate", "emit_signed_leaf_set",
         "submit_signed_leaf_set", "raw", "seal", "sealed")
for p in files:
    text = p.read_text()
    hits = [t for t in terms if t in text.lower()]
    if hits:
        print(f"{p}: {', '.join(hits)}")
PY

Repository: BaseIntelligence/base

Length of output: 4807

🏁 Script executed:

#!/bin/bash
set -eu

printf '%s\n' '--- orchestrator emitter methods ---'
sed -n '700,930p' crates/design-challenge/src/orchestrator.rs
sed -n '930,1110p' crates/design-challenge/src/orchestrator.rs

printf '%s\n' '--- submit API ---'
sed -n '1,340p' crates/challenge-common/src/submit.rs

printf '%s\n' '--- leaf emit API ---'
sed -n '1,180p' crates/challenge-common/src/leaf_emit.rs

printf '%s\n' '--- sealer relevant symbols ---'
rg -n -C 6 \
  'pub fn|pub async fn|Router|route|seal|sealed|raw|weight|#\[test\]|#\[tokio::test\]' \
  crates/gateway/src/sealer.rs crates/gateway/src/api.rs crates/gateway/tests 2>/dev/null | sed -n '1,500p'

printf '%s\n' '--- all test source flow markers ---'
rg -n -C 2 \
  'emit_signed_leaf_set|submit_signed_leaf_set|sealed\s*[:=]|sealed.*true|seal_epoch|seal|raw_weights|raw weight|failure probe|failure_probe|challenge.*valid|validate.*challenge' \
  crates/*/tests crates/*/src 2>/dev/null | sed -n '1,500p'

printf '%s\n' '--- source-level verifier ---'
python3 - <<'PY'
from pathlib import Path
roots = [Path("crates/design-challenge"), Path("crates/challenge-common"), Path("crates/gateway")]
files = [p for root in roots for p in root.rglob("*.rs")]
required = {
    "intake": ("insert_harness", "insert_run", "GatewayClient"),
    "failure probes": ("failure", "reject", "invalid"),
    "challenge validation": ("validate_bundle", "challenge_id", "verify"),
    "leaf emission": ("emit_signed_leaf_set", "make_signed_leaf"),
    "raw weight submission": ("submit_signed_leaf_set", "RawWeight"),
    "sealing": ("seal", "sealed"),
    "sealed true assertion": ("sealed: true", "sealed == true", "assert!(sealed"),
}
for p in files:
    text = p.read_text()
    lower = text.lower()
    hits = [name for name, needles in required.items()
            if any(n.lower() in lower for n in needles)]
    if hits:
        print(f"{p}: {', '.join(hits)}")
PY

Repository: BaseIntelligence/base

Length of output: 50377

🏁 Script executed:

#!/bin/bash
set -eu

printf '%s\n' '--- gateway files ---'
git ls-files 'crates/gateway' | sed -n '1,240p'

printf '%s\n' '--- seal and raw-weight endpoints across gateway ---'
rg -n -C 5 \
  'seal_epoch|admin.*seal|/v1/admin|weights/raw|post_raw|RawWeight|sealed|build_latest|build_burn_fallback' \
  crates/gateway --glob '*.rs' | sed -n '1,620p'

printf '%s\n' '--- exact flow assertions in tests ---'
rg -n -C 5 \
  '#\[test\]|#\[tokio::test\]|assert[^;]*(202|409|sealed|true|raw|seal)|failure|probe|invalid|reject|challenge_id|emit_signed_leaf_set|submit_signed_leaf_set' \
  crates/gateway crates/challenge-common crates/design-challenge \
  --glob '*test*.rs' --glob 'tests/*.rs' --glob '*.rs' | sed -n '1,700p'

printf '%s\n' '--- focused source-level test verifier ---'
python3 - <<'PY'
from pathlib import Path
roots = [Path("crates/gateway"), Path("crates/challenge-common"), Path("crates/design-challenge")]
for p in sorted({p for root in roots for p in root.rglob("*.rs")}):
    text = p.read_text()
    if "#[test]" not in text and "#[tokio::test]" not in text:
        continue
    low = text.lower()
    groups = {
        "intake": ["insert_harness", "insert_run", "post(\"/v1", "router"],
        "failure probes": ["failure", "probe", "invalid", "reject"],
        "challenge validation": ["validate_bundle", "challenge_id", "verify"],
        "leaf emission": ["emit_signed_leaf_set", "make_signed_leaf"],
        "raw submission": ["submit_signed_leaf_set", "/v1/weights/raw"],
        "sealing": ["seal_epoch", "admin/seal", "build_sealed_bundle"],
        "sealed assertion": ["sealed: true", '"sealed": true', "sealed == true"],
    }
    present = [name for name, needles in groups.items()
               if any(n.lower() in low for n in needles)]
    if present:
        print(f"{p}: {'; '.join(present)}")
PY

Repository: BaseIntelligence/base

Length of output: 50377

Add an end-to-end challenge submission test.

These tests assert only DesignEmitPlan fields. Add integration coverage for both cold-start and bounded-recovery plans through intake, failure probes, challenge validation, exact-E leaf emission, POST /v1/weights/raw, sealing, and GET /v1/weights/latest with sealed: true. Existing tests stop at sandbox execution, sanitization, and scoring.

📍 Affects 2 files
  • crates/design-challenge-task/src/emit.rs#L103-L120 (this comment)
  • crates/design-challenge/src/lib.rs#L87-L97
🤖 Prompt for AI Agents 
Verify each finding against current code. Fix only still-valid issues, skip the
rest with a brief reason, keep changes minimal, and validate.

In `@crates/design-challenge-task/src/emit.rs` around lines 103 - 120, The
emit-plan tests in crates/design-challenge-task/src/emit.rs:103-120 and the
related challenge test coverage in crates/design-challenge/src/lib.rs:87-97 need
end-to-end submission tests. Add integration coverage for both cold-start and
bounded-recovery plans, exercising intake, failure probes, challenge validation,
exact-E leaf emission, POST /v1/weights/raw, sealing, and GET
/v1/weights/latest, asserting the latest response has sealed: true; retain the
existing plan-field assertions.

Source: Coding guidelines


#[test]
fn emit_plan_noop_when_already_emitted_current() {
assert!(design_emit_plan(11, 11, 350, 360, 1000).is_none());
Expand Down
12 changes: 12 additions & 0 deletions crates/design-challenge/src/lib.rs
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -84,6 +84,18 @@ mod tests {
assert_eq!(p.pin_block, 8_815_687 - (24423 - 24413) * 360);
}

#[test]
fn emit_plan_cold_start_emits_current_immediately() {
let p = design_emit_plan(0, 24424, 10, 360, 8_816_047).unwrap();
assert_eq!(
p,
DesignEmitPlan {
epoch: 24424,
pin_block: 8_816_047
}
);
}

#[test]
fn emit_plan_noop_when_already_emitted_current() {
assert!(design_emit_plan(11, 11, 350, 360, 1000).is_none());
Expand Down
Loading