Native: move hardforks management to native Policy by cschuchardt88 · Pull Request #4726 · neo-project/neo

cschuchardt88 · 2026-08-09T15:42:37Z

codecov · 2026-08-09T15:45:14Z

Codecov Report

❌ Patch coverage is 92.48120% with 10 lines in your changes missing coverage. Please review.
✅ Project coverage is 83.94%. Comparing base (99618f4) to head (805ca59).

Files with missing lines	Patch %	Lines
src/Neo/SmartContract/Native/PolicyContract.cs	93.15%	3 Missing and 2 partials ⚠️
src/Neo/Ledger/Blockchain.cs	0.00%	2 Missing ⚠️
src/Neo/Hardfork.cs	94.11%	0 Missing and 1 partial ⚠️
...rc/Neo/SmartContract/ApplicationEngine.Contract.cs	66.66%	0 Missing and 1 partial ⚠️
src/Neo/SmartContract/Native/Notary.cs	0.00%	0 Missing and 1 partial ⚠️

Additional details and impacted files

@@             Coverage Diff             @@
##           master-n3    #4726    +/-   ##
===========================================
  Coverage      83.93%   83.94%            
===========================================
  Files            240      242     +2     
  Lines          17129    17233   +104     
  Branches        2453     2481    +28     
===========================================
+ Hits           14378    14466    +88     
- Misses          1964     1974    +10     
- Partials         787      793     +6

☔ View full report in Codecov by Harness.
📢 Have feedback on the report? Share it here.

🚀 New features to boost your workflow:

❄️ Test Analytics: Detect flaky tests, report on failures, and find test suite problems.

cschuchardt88 · 2026-08-09T15:58:46Z

Heads up @roman-khimov @AnnaShaleva — this implements #4580 (committee Policy hardfork activation), including the override-safety pieces from the discussion: public-network lock (MainNet magic + setPublicNetwork), private-only HardforkDebugOverrides for neo-express/dev, and HF_Iara as the first post-Huyao Policy-activatable hardfork. Review welcome when you have time.

Add Policy.enableHardfork/getHardfork (HF_Huyao) so post-Huyao hardforks activate from the next block after a committee-signed call. Wire IsHardforkEnabled, native init, and contract state to honor on-chain Policy heights; unknown hardfork ids fail the block so outdated nodes stop until upgraded.

Address neo#4580 comment on override safety: well-known MainNet magic and on-chain setPublicNetwork seal public chains so post-Huyao activation is Policy-only. Add HardforkDebugOverrides for neo-express/private nets, HF_Ifrit as the first Policy-activatable hardfork, config divergence detection, and tests for misconfiguration failure modes.

Cover private-net enable rejections, public-config issue reporting, HardforkDebugOverrides validation/load paths, and Policy-aware NativeContract init/active helpers after rebasing onto HF_Iara.

AnnaShaleva · 2026-08-10T08:27:35Z

@cschuchardt88 the final design of this feature is discussed with @roman-khimov and presented in nspcc-dev/neo-go#4377. So we need C# version to follow this PR. I haven't yet compared your implementation with nspcc-dev/neo-go#4377, will review deeply later.

cschuchardt88 · 2026-08-10T09:22:55Z

@AnnaShaleva can you check #4580 (comment) and give me some insights my concerns.

Address Anna's review on neo#4726 / nspcc-dev/neo-go#4377: - Rename enableHardfork/getHardfork to activateHardfork/getHardforkActivationHeight - Use raw hardfork names (Iara) and return Null when unset - Rename HardforkEnabled to HardforkActivationScheduled - UnknownHardforkException after AssertCommittee so outdated nodes halt - Drop public-network marker and HardforkDebugOverrides; post-Huyao is Policy-only - Require a snapshot in NativeContract.IsActive

cschuchardt88 · 2026-08-20T21:53:10Z

Thanks @AnnaShaleva — this push aligns the C# side with your review and nspcc-dev/neo-go#4377.

Addressed:

IsActive now requires a snapshot (no null-passing overload).
Dropped Prefix_PublicNetwork / setPublicNetwork / getPublicNetwork — only hardfork heights are stored.
Event renamed to HardforkActivationScheduled; hardfork payload is the raw string (Iara), not an int.
Removed the #region.
getHardfork → getHardforkActivationHeight; enableHardfork → activateHardfork.
Hardfork parameter is a string (Iara). Unset/unknown getter returns Null instead of -1.
Unknown names throw UnknownHardforkException after AssertCommittee. Blockchain.Persist rethrows it so the node stops instead of FAULTing the tx. Random senders cannot halt the node.
Removed HardforkDebugOverrides and the public/private network split. Post-Huyao activation is Policy-only.

Not in this push (same TODO as neo-go #4377): moving Aspidochelone–Huyao activations into Policy and initializing them at Huyao. Happy to do that as a follow-up once the neo-go design for it lands.

- Parse Policy hardfork names as the exact raw string only (Iara, not iara/HF_Iara). - IsHardforkEnabledDelegate now takes settings and snapshot; drop the settings-only GetContractState overload. - Restore the omitted-hardfork-is-disabled comment in IsInitializeBlock. - activateHardfork uses engine.IsHardforkEnabled only; notify with the raw name. - Store A-H config heights in Policy on Huyao init; Policy.Activations includes Iara. - Reject post-Huyao entries in ProtocolSettings.Hardforks. - Remove DetectHardforkConfigDivergence and redundant Policy tests.

cschuchardt88 · 2026-08-25T15:01:52Z

Thanks @AnnaShaleva — this push addresses your 2026-08-25 review.

Policy names and activateHardfork

Hardfork argument is the exact raw string only (Iara). iara and HF_Iara are rejected (unknown name after committee).
Dropped the extra snapshot Contains check; engine.IsHardforkEnabled(hf) is enough.
Notification is [hardfork, activationHeight] using the input name.
Error messages use the input name (must be activated via ProtocolSettings configuration, not Policy / already enabled).
Removed the instance IsHardforkEnabled(snapshot, hf, index) overload; the static (settings, snapshot, hf, index) form is used everywhere.

NativeContract

IsHardforkEnabledDelegate now takes ProtocolSettings and IReadOnlyStore? so the snapshot dependency is explicit. Native cache uses PolicyContract.IsHardforkEnabled as the method group.
Removed GetContractState(settings, blockHeight). Remaining overload requires a snapshot.
Restored the omitted-hardfork-is-disabled comment in IsInitializeBlock.
IsActive always goes through Policy's combined checker.

Huyao init / Activations

InitializeAsync(HF_Huyao) copies config-managed A–H heights into Policy storage.
Policy.Activations is [null, HF_Iara] (same pattern as Notary), so Iara is a Policy initialize height.

ProtocolSettings

CheckingHardfork rejects post-Huyao entries (HF_Iara in config throws).
Removed DetectHardforkConfigDivergence (tests-only).
Removed Hardforks.TryParse (case-insensitive / HF_ prefix); replaced with TryParseExact.

Tests

Dropped the Policy ABI/manifest test (ContractManagement already covers it).
Dropped the local-config Iara test; added Load-rejects-Iara and case-sensitive name tests.
Huyao init test asserts A–H heights are stored.
IsInitializeBlock at a Policy-stored Iara height is true via Activations.

78 related unit tests passed (UT_PolicyContract, UT_NativeContract, UT_ProtocolSettings).

After Huyao initialize copies Aspidochelone-Huyao heights into Policy, TryGetActivationHeight prefers those stored heights so IsActive is Policy-backed (neo-go#4377). Config is used only until Huyao is enabled.

cschuchardt88 · 2026-08-28T02:14:32Z

Completed the remaining neo-go#4377 / Anna TODO on this PR: use Policy as the runtime source of truth for Aspidochelone–Huyao once Huyao is enabled.

Huyao InitializeAsync already copied A–H heights into Policy storage. TryGetActivationHeight now:

takes the block index
for post-Huyao forks: Policy only
for A–H: Policy after Huyao is enabled in config (the copy exists); ProtocolSettings.Hardforks until then

So IsActive / IsInitializeBlock stop consulting local config for A–H after Huyao, matching Roman’s “initialize A–H at Huyao, then keep only Policy in IsActive.” getHardforkActivationHeight("Aspidochelone") also returns the stored height.

Tests: Check_AfterHuyao_ConfigManagedHeightsComeFromPolicy (Policy wins over a mismatched local Faun height), plus Huyao init / initialize-block coverage. 79 related tests passed.

Not in this PR: the same A–H copy inside nspcc-dev/neo-go#4377, and committee multi-sig packaging for activateHardfork txs.

Cover getHardforkActivationHeight after activate, missing Policy height, activate without a persisting block, UnknownHardforkException wrapping, and Hardforks.TryParseExact/GetName.

vncoelho · 2026-09-21T14:17:20Z

Yes, @AnnaShaleva
Thanks for this update.

But my main concern is regard the previous conducted tests . If anyone has already tested...anyway.

The PR looks good. I am finishing some basic testing.+

shargon · 2026-09-29T14:34:16Z

+            if (engine.SnapshotCache.Contains(key))
+                throw new InvalidOperationException($"Hardfork {hardfork} is already scheduled.");
+
+            uint activationHeight = checked(engine.PersistingBlock.Index + 1);


Maybe we should send the num of blocks after activation, to have margin between sign, relay, and fork

Can be done, but why do we need it? We don't have a mechanism of scheduled hardfork deactivation anyway.

send the num of blocks

Do you want it to be a parameter of ActivateHardfork method?

Preparation, you can publish that fork will be in specific height, like this is not possible to know when will be active

To me it's pretty useless, I'd rather have it activated once transaction gets in, it's just easier to handle. We don't know when block X will happen anyway.

Contracts can be prepared, with this new logic you are no only changing how hardfork will be activated, you are changing how and when the users will know it

Maybe we should send the num of blocks after activation, to have margin between sign, relay, and fork

Or we set the number of blocks after agreement when required number of signatures is reached (but remember that this will be prune to interpretation of txs and should be ensured by onchain TXs only).

Or, at least, we need a fixed height upon agreement, if not reached agreement until that we try again. That is the best model to me. Fixed height and tentative until it.

Preparation, you can publish that fork will be in specific height, like this is not possible to know when will be active

My idea is the same, @shargon

To me it's pretty useless, I'd rather have it activated once transaction gets in, it's just easier to handle. We don't know when block X will happen anyway.

I do not think so, to much uncertainties, too much prone to manipulation. Fixed height and fixed rule with a Preparation.

Contracts can be prepared, with this new logic you are no only changing how hardfork will be activated, you are changing how and when the users will know it

I am fully aligned with you, @shargon

Implemented, @shargon let's check the updated version. We also added an ability to reschedule the already scheduled hardfork. It's useful in case if hardfork is already scheduled but some bug is discovered and a fix should be applied prior to hardfork activation.

shargon · 2026-09-29T14:36:45Z

+                return false;
+            }
+
+            if (settings.IsHardforkEnabled(ProtocolSettings.LastConfigManagedHardfork, currentIndex)


Why check of is enabled? You want to know the activation height

Because starting from LastConfigManagedHardfork we store height in the Policy's storage. So if LastConfigManagedHardfork is enabled then we should refer to the Policy instead of config.

Signed-off-by: Anna Shaleva <shaleva.ann@nspcc.ru>

shargon · 2026-10-01T06:09:00Z

                return false;

-            foreach (Hardfork value in Enum.GetValues<Hardfork>())
+            if (Enum.TryParse(typeof(Hardfork), "HF_" + name, false, out var fork) && Enum.GetNames<Hardfork>().Any(n => n.Equals("HF_" + name, StringComparison.Ordinal)))


Is not enough with tryParse?

We need to check that the resulting enum is defined. But you're right, we add HF_ prefix to name, so any non-valid name will fail parsing.

Signed-off-by: Anna Shaleva <shaleva.ann@nspcc.ru>

AnnaShaleva · 2026-10-05T08:53:02Z

Updated to the latest master.

[Flags] lets Enum.TryParse accept comma-separated aliases that equal a named value. TryParseExact now requires the canonical HF_ name, so activateHardfork only accepts exact names such as Iara. The macOS Test job failed UT_TaskManager.InvalidBlock_AbortsThePeerThatSuppliedIt because peer.Send returned before the mailbox recorded the hash. Drive the tests that assert ReceivedBlockHashes immediately through TestActorRef.Receive. Ref. neo-project#4726

cschuchardt88 · 2026-10-06T00:48:07Z

Pushed 72039f2c. Ubuntu, Windows, and macOS Test are green.

Hardfork name parsing. Hardfork is [Flags], so Enum.TryParse accepts comma-separated aliases that collapse onto a named value (Basilisk, HF_Cockatrice equals HF_Domovoi). Hardforks.TryParseExact now requires the canonical HF_ name. activateHardfork still takes the exact raw string (Iara). iara, HF_Iara, and flag aliases are rejected. Covered in Check_Hardforks_TryParseExact_AndGetName.

macOS TaskManager failure. UT_TaskManager.InvalidBlock_AbortsThePeerThatSuppliedIt failed on macos-latest because peer.Send returned before the actor recorded the block hash. That test and the siblings that assert ReceivedBlockHashes immediately now use TestActorRef.Receive, so the actor runs on the test thread. No production TaskManager change.

activateHardfork and getHardforkActivationHeight now take ByteArray (0x12) instead of String (0x13) so the Huyao Policy manifest matches nspcc-dev/neo-go#4377. Names are still UTF-8 canonical HF names parsed by Hardforks.TryParseExact. Ref. neo-project#4726

This reverts commit 640f1f6.

AnnaShaleva · 2026-10-06T08:38:00Z

640f1f6 is reverted, we need a fully-qualified String here. NeoGo implementation is out-of-date and will be aligned with the C# core once this PR is finalized.

Ref. neo-project#4726 (comment). Signed-off-by: Anna Shaleva <shaleva.ann@nspcc.ru>

AnnaShaleva · 2026-10-06T10:36:14Z

@neo-project/core let's review the updated API one more time.

cschuchardt88 · 2026-10-06T13:21:20Z

+        }
+
+        [TestMethod]
+        public void Load_IgnoresHardforkDebugOverrides_AndDoesNotEnableIara()


This test is no longer need. It will always fail.

Agree, removed.

Ref. neo-project#4726 (comment). Signed-off-by: Anna Shaleva <shaleva.ann@nspcc.ru>

AnnaShaleva · 2026-10-08T11:09:19Z

This PR slightly depends on #4764, so let's wait for the #4764 to settle and I'll update it afterwards.

github-actions Bot added the N3 label Aug 9, 2026

cschuchardt88 added 3 commits August 9, 2026 18:55

test(Policy): expand hardfork activation coverage for codecov

08f3acf

Cover private-net enable rejections, public-config issue reporting, HardforkDebugOverrides validation/load paths, and Policy-aware NativeContract init/active helpers after rebasing onto HF_Iara.

cschuchardt88 force-pushed the fix/hardfork-policy-activation branch from b8f5763 to 08f3acf Compare August 9, 2026 23:00

AnnaShaleva reviewed Aug 13, 2026

View reviewed changes

AnnaShaleva added the Hardfork label Aug 13, 2026

AnnaShaleva mentioned this pull request Aug 19, 2026

[2026-08-20] 🗓️ Core Dev Regular Meeting Discussion #4728

Closed

cschuchardt88 changed the title ~~feat(Policy): enable hardforks via committee transaction (neo#4580)~~ feat(Policy): activate hardforks via committee transaction (neo#4580) Aug 20, 2026

AnnaShaleva reviewed Aug 25, 2026

View reviewed changes

vncoelho and others added 2 commits August 25, 2026 08:54

Merge branch 'master-n3' into fix/hardfork-policy-activation

b39f2a6

cschuchardt88 mentioned this pull request Aug 25, 2026

Native: introduce TemporaryStorage native contract #4727

Merged

10 of 18 tasks

cschuchardt88 added 2 commits August 27, 2026 21:44

Merge branch 'master-n3' into fix/hardfork-policy-activation

ea04f51

Use Policy for A-H hardfork heights after Huyao (neo#4726)

85d9996

After Huyao initialize copies Aspidochelone-Huyao heights into Policy, TryGetActivationHeight prefers those stored heights so IsActive is Policy-backed (neo-go#4377). Config is used only until Huyao is enabled.

cschuchardt88 mentioned this pull request Aug 28, 2026

Map hardforks onto neo-vm VmFeatures (dpends on neo-vm#592) #4740

Closed

cschuchardt88 added 2 commits August 28, 2026 04:25

Add Policy hardfork helper coverage

3ce6f9d

Cover getHardforkActivationHeight after activate, missing Policy height, activate without a persisting block, UnknownHardforkException wrapping, and Hardforks.TryParseExact/GetName.

Merge branch 'master-n3' into fix/hardfork-policy-activation

566b2ef

cschuchardt88 mentioned this pull request Aug 28, 2026

Review tracker: VmFeatures stack, neo-node parse/oracle/tui #4744

Open

3 of 17 tasks

AnnaShaleva reviewed Aug 28, 2026

View reviewed changes

AnnaShaleva force-pushed the fix/hardfork-policy-activation branch from 04df42d to b69df3a Compare September 15, 2026 11:50

AnnaShaleva changed the title ~~feat(Policy): activate hardforks via committee transaction (neo#4580)~~ Policy: move hardforks management to native Policy Sep 15, 2026

AnnaShaleva requested a review from superboyiii September 21, 2026 14:14

AnnaShaleva and others added 2 commits September 25, 2026 11:43

Merge branch 'master-n3' into fix/hardfork-policy-activation

0b49340

Merge branch 'master-n3' into fix/hardfork-policy-activation

83cfdc2

shargon reviewed Sep 27, 2026

View reviewed changes

Comment thread src/Neo/Ledger/Blockchain.cs

AnnaShaleva requested a review from shargon September 28, 2026 09:09

shargon reviewed Sep 29, 2026

View reviewed changes

AnnaShaleva added 2 commits September 30, 2026 18:11

Mark Hardfork with Flags attribute

7173620

Signed-off-by: Anna Shaleva <shaleva.ann@nspcc.ru>

Update Hardforks.TryParseExact logic

056ea73

Signed-off-by: Anna Shaleva <shaleva.ann@nspcc.ru>

AnnaShaleva requested a review from shargon September 30, 2026 17:35

shargon reviewed Oct 1, 2026

View reviewed changes

Simplify Hardforks.TryParseExact

276e7eb

Signed-off-by: Anna Shaleva <shaleva.ann@nspcc.ru>

AnnaShaleva requested review from Jim8y, Wi1l-B0t and shargon October 1, 2026 08:13

shargon mentioned this pull request Oct 1, 2026

[🗓️ 2026-10-01] Core Dev Regular Meeting Discussion #4758

Closed

Merge branch 'master-n3' into fix/hardfork-policy-activation

bc16273

cschuchardt88 and others added 2 commits October 5, 2026 23:36

Revert "Use ByteArray ABI for Policy hardfork methods"

9c22d69

This reverts commit 640f1f6.

Add activationDelay parameter to Policy's activateHardfork

22d8d41

Ref. neo-project#4726 (comment). Signed-off-by: Anna Shaleva <shaleva.ann@nspcc.ru>

cschuchardt88 commented Oct 6, 2026

View reviewed changes

AnnaShaleva mentioned this pull request Oct 8, 2026

ApplicationEngine: fix IsHardforkEnabled behaviour #4764

Open

10 of 19 tasks

AnnaShaleva added 2 commits October 8, 2026 14:02

Remove stale unit-test

bfef3da

Ref. neo-project#4726 (comment). Signed-off-by: Anna Shaleva <shaleva.ann@nspcc.ru>

Merge branch 'master-n3' into fix/hardfork-policy-activation

805ca59

Conversation

cschuchardt88 commented Aug 9, 2026 • edited by AnnaShaleva Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

codecov Bot commented Aug 9, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Codecov Report

Uh oh!

cschuchardt88 commented Aug 9, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

AnnaShaleva commented Aug 10, 2026

Uh oh!

cschuchardt88 commented Aug 10, 2026

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

cschuchardt88 commented Aug 20, 2026

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

cschuchardt88 commented Aug 25, 2026

Uh oh!

cschuchardt88 commented Aug 28, 2026

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

vncoelho commented Sep 21, 2026

Uh oh!

Uh oh!

Choose a reason for hiding this comment

Uh oh!

AnnaShaleva Sep 30, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Choose a reason for hiding this comment

Uh oh!

shargon Oct 1, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

cschuchardt88 commented Aug 9, 2026 •

edited by AnnaShaleva

Loading

codecov Bot commented Aug 9, 2026 •

edited

Loading

cschuchardt88 commented Aug 9, 2026 •

edited

Loading

AnnaShaleva Sep 30, 2026 •

edited

Loading

shargon Oct 1, 2026 •

edited

Loading