docs(test13): hardfork RC series (review-only against chain/gnoland1)#5589
Closed
aeddi wants to merge 92 commits into
Closed
docs(test13): hardfork RC series (review-only against chain/gnoland1)#5589aeddi wants to merge 92 commits into
chain/gnoland1)#5589aeddi wants to merge 92 commits into
Conversation
PR gnolang#5486 commit bd3580d ("refactor: absorb misc/hardfork into 'gnogenesis fork' subcommand") moved misc/hardfork/ into contribs/gnogenesis/internal/fork/ and rewired the CLI: misc/hardfork test → gnogenesis fork test That refactor updated misc/deployments/gnoland-1/generate-genesis.sh but missed the references in misc/hf-glue/. As a result, scripts/replay-log.sh (invoked by `make replay-log` and `make reports`) still does: cd "$REPO/misc/hardfork" go run . test ... which fails with "No such file or directory" because misc/hardfork no longer exists on the PR head. Update the cd target to contribs/gnogenesis and the subcommand to `fork test` to match the new layout. Other stale refs in misc/hf-glue/ (Makefile smoketest target, fetch-from-dir.sh, README, lib/hf.sh comment) are also broken but addressed in separate commits / left as docs.
scripts/report-replay.sh runs under `set -euo pipefail`. The assignment:
missing_imports=$(grep -oE 'could not import gno\.land/[^ \"\\]+' "$tmp" \
| sort | uniq -c | sort -rn | head -20)
fails the whole script when the [FAIL] log contains no "could not import"
lines: grep exits 1, pipefail propagates the non-zero status, and the
outer command substitution triggers set -e.
This is the happy path when a hardfork replay has no cascade failures —
e.g. the current run on chain/test13-base has exactly one root-cause
failure (r/sys/txfees storage deposit) and zero cascades, so the pipeline
returned empty and aborted the script before the report was written.
Append `|| true` to the pipeline so an empty result is treated as
"missing_imports is empty", consistent with how the adjacent
root_cause_fails assignment already handles it.
scripts/report-replay.sh runs under `set -u`. The line:
echo "_Generated $(date ...) from $LOG_"
is parsed as expanding a variable named `LOG_` (trailing underscore is a
valid identifier character), not `LOG` followed by a literal underscore.
Since $LOG_ is unbound, set -u aborts the script with:
line 58: LOG_: unbound variable
The trailing underscore was intended as the closing italic marker for
the markdown header `_Generated ... from /path/to/replay.log_`.
Wrap the variable in braces — `${LOG}_` — so bash parses it as $LOG
followed by a literal underscore, preserving the markdown emphasis.
… failures Root cause ---------- master's storage-deposit code (4e1745a, gnolang#5415) locks a deposit from msg.Creator for every addpkg, proportional to realm size. This feature did not exist on gnoland1. The 85 genesis-mode addpkg txs in the gnoland1 snapshot were all signed with max_deposit="" (the field did not exist yet). Under the new SDK the empty field defaults to the parameter ceiling (600 Mugnot), so the actual transfer is `diff × 100 ugnot` per realm. The creator g1r929wt2qplfawe4lvqv9zuwfdcz4vxdun7qh8l deploys 7 consecutive r/sys/* realms in genesis-mode, depleting its 58.7 Mugnot balance. The 8th deploy (r/sys/txfees) then fails with insufficient funds. Every downstream realm that imports r/sys/txfees cascades. Why cherry-pick is not viable ------------------------------ The realm code on master is byte-for-byte identical to gnoland1's. The gap is purely between old unsigned tx format and new SDK semantics. There is nothing to cherry-pick: the fix would have to be a new feature (e.g. a genesis-mode VM bypass), which is a larger PR outside this testbed's scope. --skip-genesis-sig-verification is already set; leaving the payload stable is the honest choice. Design: hf_topup_balance ------------------------- New DSL primitive in lib/hf.sh. Called from migrate.sh after the realm patches section. Applied in a post-assembly Python pass that: • opens $OUT/genesis.json • increments the target address's balance line (or appends if absent) • writes the mutated genesis back in place • writes $OUT/TOPUP-REPORT.md — a persistent audit table with before / after / delta / reason for every synthetic change The Python path uses an indexed O(1) lookup (dict keyed by addr+denom) because app_state.balances is a multi-million-entry list on mainnet-scale snapshots and a linear scan would be too slow. After the pass, hf_assemble prints a red `hf_warn` banner: "N synthetic balance top-up(s) applied — see $OUT/TOPUP-REPORT.md" This is intentionally noisy: synthetic state must not be silent. Visibility in STATE-REPORT.md ------------------------------ check-state.sh now reads $OUT/TOPUP-REPORT.md (when present) and emits a "⚠ Synthetic state modifications" table that probes the topped-up addresses on both local and prod chains. A reviewer looking at STATE-REPORT.md sees the full divergence picture: local/prod deltas for these addresses are expected and documented, not replay bugs. Files changed ------------- scripts/lib/hf.sh — new _HF_TOPUPS state var, hf_warn helper, hf_topup_balance DSL function, _hf_apply_topups internal, hf_assemble calls _hf_apply_topups scripts/migrate.sh — section 3b calls hf_topup_balance for the r/sys/* creator with a 1 Bugnot top-up and a documented reason scripts/check-state.sh — synthetic modifications section in STATE-REPORT Result: gnogenesis fork test now reports 2696/2696 OK, 0 failures.
alloc per-byte, first step to correct/improve gas assumption for memory allocation. (cherry picked from commit 5d5f921)
Safari iOS does not re-evaluate CSS custom properties in SVG inline style attributes when variable values change dynamically. The logo fills were invisible in dark mode because `fill: var(--s-logo-hat)` set via inline style was never recalculated after the theme switch. **Fix: moved fills to CSS stylesheet rules (`b-gnome/b-logo blocks`) so the browser cascades correctly on theme change.** Tested on Safari iOS (iPhone) in dark mode: logo now renders correctly on initial load without requiring any repaint. <img width="45%" height="1434" alt="IMG_4054" src="https://github.com/user-attachments/assets/91dcc5e4-17b0-4f9c-8baa-84467185d390" /> <img width="45%" height="1434" alt="IMG_4053" src="https://github.com/user-attachments/assets/2048fc94-71d6-458d-9de6-325c659b4264" /> (cherry picked from commit c1a785a)
…gnolang#5251) Fix handshake validation to check the received peer NodeInfo (not local NodeInfo), closing a path where malformed peer payloads could bypass validation and trigger nil dereference later in handshake flow. Also harden NodeInfo.Validate by rejecting nil NetAddress explicitly. Tests: - add NodeInfo.Validate regression for missing NetAddress - add transport handshake regression for malformed peer node info Context: NEWTENDG-161 (cherry picked from commit 553b1c7)
…ges (gnolang#5305) - Static markdown pages (configured via `--aliases=/path=static:file.md`) rendered **Source** and **Actions** header buttons that returned errors when clicked, since there is no realm backing the content. - Add a `Static` bool field to `HeaderData` that suppresses Source/Actions links, keeping only the Content button for static markdown pages. - Homepage general links (About, Docs, GitHub) remain unaffected. (cherry picked from commit 7f6f798)
Co-authored-by: Morgan Bazalgette <morgan@morganbaz.com> (cherry picked from commit cfe24df)
- Add compile-time validation of map key comparability - Add compile-time detection of duplicate constant keys in map literals (cherry picked from commit eef3b22)
fix gnolang#5270 & dup. gnolang#5273 --- It implements abstain vote in the gov/dao, it was already showned in the UI but the business logic vote did not handle this kind of vote, i also opened a PR with an alternative that remove the Abstain vote from the UI (since the impl. does not handle it): gnolang#5272 But this version is better IMO, abstain vote can be important >⚠️ Also, Should abstain votes reduce the effective denominator for supermajority? See this msg for more details: gnolang#5271 (comment) --------- Co-authored-by: Morgan <morgan@morganbaz.com> (cherry picked from commit ce2ba67)
Plain byte-index assignment (bs[i] = v) on Data-backed []byte slices was silently dropping mutations across transactions. PointerValue.Assign2 returned early for DataByteType after calling SetDataByte, never reaching the rlm.DidUpdate call that marks the backing ArrayValue dirty. The fix mirrors the pattern already used in op_inc_dec.go for data[i]++. Adds zrealm_databyte0.gno filetest verifying the Realm diff shows the backing Data array is now persisted after plain index assignment. Reported-by: torxeon (HackenProof NEWTENDG-98) Co-authored-by: ltzmaxwell <ltz.maxwell@gmail.com> (cherry picked from commit a3a356e)
fix this: https://github.com/gnolang/gno/actions/runs/17917650414/job/50943950259?pr=4040 by properly allocate while restoring objects from storage. see comments in files. --------- Co-authored-by: Morgan <morgan@morganbaz.com> (cherry picked from commit e6da902)
…lang#5037) This PR introduces "protected" Boards2 realm functions that can only be called from sub-realms deployed under the `gno.land/r/gnoland/boards2` namespace. The "protected" functions are necessary to implement, or move, specific board features to sub-realms. They could also be useful, for example, to support migrating data between versions. It also adds a new `hub` sub-realm with safe types and public functions that can be called from other realms or off-chain apps to read Boards2 realm data. This realm initially allows Boards mobile app to better integrate with Boards2 realm by being able to read boards data. Moving or implementing new features as sub-realms would help keeping Boards2 realm leaner and would potentially allow for a more "micro services" oriented approach for new features. --------- Co-authored-by: Rémi BARBERO <d4ryl00@gmail.com> (cherry picked from commit 4d4478b)
GC and MemStats are only used in filetests (gnovm/tests/files/alloc_*.gno) and have no legitimate use in on-chain code. Keeping them in the production stdlibs would lock in implementation details — specifically, the allocator's internal byte-counting format — as part of the deterministic on-chain execution environment. Moving them to gnovm/tests/stdlibs/runtime/ ensures they are only available during testing and cannot be imported by realm or package code deployed on-chain. BREAKING CHANGE: Backward incompatible with gnoland1. (cherry picked from commit afd7e48)
…gnolang#5280) Boards2 is now initialized with a single address which is a multisig address (`g1rp7cmetn27eqlpjpc4vuusf8kaj746tysc0qgh`). This PR removes address `g16jpf0puufcpcjkph5nxueec8etpcldz7zwgydq` from the initial list leaving only the multisig one. The removed address has been replaced by the multisig in all tests. (cherry picked from commit 4a84645)
Fix this HP issue: https://dashboard.hackenproof.com/manager/companies/newtendermint/gno-dot-land/reports/NEWTENDG-164 --- This PR does 3 things: - Fix int64() cast that caused large uint64 constants to bypass overflow validation when converting to smaller types - Improve overflow error message to match Go's format - allow int -> string at runtime, was already allowed at preprocess since convertTo at preprocess is called only on numeric target (cherry picked from commit 6a6fc4c)
…gnolang#5315) Audit and reclassify logger calls throughout the tm2, gno.land, and contribs layers: - Info → Debug: per-operation noise (mempool tx accept/reject/recheck, RPC request/response per call, ABCI queries, consensus fast-sync message filtering, consensus replay step messages, p2p connection message details) - Info → Warn: anomalous conditions that indicate protocol violations or backpressure (unexpected block from peer, invalid peer, internal msg queue full) - Debug → Info: significant lifecycle events that operators need visibility into (p2p listener closed, GnoVM packages preprocessed with elapsed time) - Remove per-packet Debug logs in MConnection (Send, TrySend, Received bytes, Read PacketMsg) that are too noisy even at debug - Fix switch dial loop to skip logging "dialing peer" for already- connected peers (log was before the duplicate-check) - Improve --log-level flag help text to list valid values This should greatly improve the number of noisy logs from the node. (cherry picked from commit a82373a)
gnolang#5228) `SetPeerHeight` only ever raised `pool.maxPeerHeight`. If a peer announced a high fake height and then corrected it, the pool's max stayed permanently stuck, causing `IsCaughtUp()` to return `false` forever and preventing the node from leaving fast-sync mode. This PR fixes it by adding an `else if` branch in `SetPeerHeight` that calls the existing `updateMaxPeerHeight()` when a peer's new height is below the current maximum. (cherry picked from commit 8d17f08)
closes gnolang#1611 --------- Co-authored-by: Morgan <morgan@morganbaz.com> (cherry picked from commit 908d1f7)
fix: gnolang#5335 I've updated the static file and the home realm, and I assume the blog post are gonna be uploaded later. I've fixed the incorrect event redirection. For the about and help, we should deploy the `r/gnoland/home` realm (defined in `gno.land/pkg/gnoweb/app.go`), or we have to modify gnoweb for this pre existing bug. In my opinion, this static file provided by default should be as generic as possible, which is why I deleted the different sections. Do we want to keep `"gno.land/p/leon/svgbtn"`? (To reduce dependency) Co-authored-by: moul <94029+moul@users.noreply.github.com> (cherry picked from commit fe7e38b)
## Summary - Add static markdown pages for all broken alias links on gno.land (`/about`, `/gnolang`, `/ecosystem`, `/start`, `/license`, `/contribute`, `/links`, `/partners`, `/events`) - Content ported from the realm source code in `examples/gno.land/r/gnoland/pages/` and staging events - Update `home-alias` docker-compose to serve static pages via `--aliases` flag - Remove dead `/r/docs/home` link from homepage template - Fix `/r/devrels/events` → `/events` alias path in homepage template ## Context The current gno.land homepage has many dead links pointing to `/r/gnoland/pages:p/*` realms that exist on staging but not on mainnet. This PR creates static markdown files and configures gnoweb aliases to serve them, restoring all broken pages. ### Pages added | Path | Content | |------|---------| | `/about` | Platform overview | | `/gnolang` | Gno language description | | `/ecosystem` | Ecosystem projects | | `/start` | Getting started | | `/license` | GPL license text | | `/contribute` | Contributor guide | | `/links` | Important links | | `/partners` | Fund & grants | | `/events` | Past events list | ## Test plan - [ ] Verify docker-compose starts gnoweb with all aliases - [ ] Check each static page renders correctly - [ ] Verify homepage links resolve (no more 404s) - [ ] Run `gno test` on home package to verify filetest passes 🤖 Generated with [Claude Code](https://claude.com/claude-code) --------- Signed-off-by: moul <94029+moul@users.noreply.github.com> Co-authored-by: Nemanja Aleksic <kouteki@gmail.com> (cherry picked from commit ccbcd67)
Add a beta banner that redirects users to the GitHub issues template for feedback. ~The current text is somewhat random; we can likely find a better option.~ added a generic env flag, to easily change the banner on purpose. <img width="1271" height="606" alt="Screenshot 2026-03-19 at 14 28 25" src="https://github.com/user-attachments/assets/ed2d2186-509d-4104-8b39-824e557a931a" /> (cherry picked from commit 3c9f07b)
Update the content on the static pages Related to gnolang#5355 (cherry picked from commit 27ee208)
…nolang#5099) Duplicate denoms in Coins is generally invalid. Currently, it simply returns the first encountered Coin, but I think the more correct behaviour is to panic so that execution doesn't continue with multiple coins of the same value. gnolang#5096 modifies NewCoins and Coins.Add to implement a behaviour I'm sure some users will expect (joining coins together). --------- Co-authored-by: David <60177543+Davphla@users.noreply.github.com> (cherry picked from commit 81d9f80)
…nolang#5249) Co-authored-by: Morgan Bazalgette <morgan@morganbaz.com> (cherry picked from commit 2d7f193)
Related with gnolang#5374 --------- Signed-off-by: moul <94029+moul@users.noreply.github.com> (cherry picked from commit 0886378)
…lang#5247) Fixes: [dashboard.hackenproof.com/manager/companies/newtendermint/gno-dot-land/reports/NEWTENDG-68](https://dashboard.hackenproof.com/manager/companies/newtendermint/gno-dot-land/reports/NEWTENDG-68). Alternative to gnolang#5153. The previous initialization algorithm used recursive depth-first traversal of variable dependencies, which could produce incorrect initialization order and was susceptible to non-determinism (map iteration over dependency sets). The Go specification mandates: > Within a package, package-level variable initialization proceeds stepwise, with each step selecting the variable earliest in declaration order which has no dependencies on uninitialized variables. ## Changes ### Initialization order rewrite **`preprocess.go`**: Replace the post-preprocessing `findDependentNames` AST walker with a dependency-tracking approach that runs as a dedicated coda pass (`codaInitOrderDeps`) after `preprocess1`: - `addDependencyToTopDecl` — called from `codaInitOrderDeps` whenever a `NameExpr` or `SelectorExpr` resolves to a package-level name in the same package. It walks up the node stack to find the enclosing top-level `ValueDecl` or `FuncDecl` and records the dependency as a `Name` in an `ATTR_DECL_DEPS` attribute on that node. Method dependencies are encoded as `"Type.Method"` so receiver bodies are tracked without re-walking the AST. - `resolveDeclDep` — given a dependency name (plain or `"Type.Method"`), looks up the corresponding `Decl` using `GetLocalIndex` + `NameSources` for correct, panic-safe `*DeclaredType` resolution. - `resolveEffectiveDeps` — memoized DFS from all declarations, following `ATTR_DECL_DEPS` edges and collapsing `FuncDecl` bodies transitively to discover indirect `*ValueDecl` dependencies. O(V+E). Detects circular variable dependencies and panics with the full chain (e.g. `circular dependency: A -> f -> B -> A`). **`machine.go`**: Replace the recursive `runDeclarationFor` + `loopfindr` approach with a two-phase iterative algorithm directly implementing the Go spec: 1. Build an ordered `[]Decl` pending list from all non-`FuncDecl` declarations across all files, preserving source declaration order. 2. Kahn's topological sort with a min-heap keyed on declaration index to always pick the earliest-in-declaration-order ready entry. O(V + E + V log V). This eliminates non-determinism from map iteration and incorrect ordering from depth-first traversal. **`nodes.go`**: Remove the now-unused `findDependentNames`, `GetExternNames`, `addExternName`, and `isFile` helpers. The `Externs` field on `StaticBlock` is retained for amino serialization backward-compatibility but is no longer populated. ### Composite literal non-const key fix **`preprocess.go`**: The old code allowed non-constant variables as slice/array composite literal keys (e.g. `[]int{a: b, c: d}` where `a` and `c` are runtime variables). This is invalid per the Go spec, which requires constant index expressions. The new code panics with `"slice/array literals may not contain non-const keys"`, matching Go's `go/types` error. **`uverse_test.go`**: `TestIssue1337PrintNilSliceAsUndefined`'s "print composite slice" case used `a, b, c, d := 1, 2, 3, 4` as composite literal keys. Updated to `const a, b, c, d = 1, 2, 3, 4` since these must be constant expressions. ### Tests - Add `var_initorder{1–26}.gno` filetests covering declaration chains, multi-name specs, blank identifiers, method-receiver dependencies (value, pointer, auto-addressed), cross-file transitive deps, cross-package method guards, interface dispatch, recursive methods, circular deps, embedded struct method promotion (value and pointer receiver), func-literal-var circular deps, multiple blank decls with side-effects, deep transitive FuncDecl→ValueDecl chains, variable shadowing in function bodies, and doubly-nested closures. - Add `var_initorder_crossfn.gno` and `var_initorder_xpkgmethod.gno` for cross-file and cross-package cases. - Add `TestInitOrderDeterminism` in `preprocess_test.go` which runs a complex dependency graph 100 times to verify stable output. - Add `TestCircDepDeterminism` which verifies circular-dependency error messages are deterministic. - Add post-Kahn's completeness assertion in `runFileDecls` that panics if any declaration was not processed (guards against missed cycles or reverse-dep notification gaps). - Update `recursive10.gno`, `recursive11.gno`, and `closure.gno` for the new circular-dependency error format. - Update `composite15.gno` expected output for the new non-const key error. ### Integration test updates `atomicswap.txtar`: Gas amounts changed slightly (e.g. 454000 → 453800) as a side effect of the new initialization order affecting the number of VM operations executed during package loading. --------- Co-authored-by: ltzmaxwell <ltz.maxwell@gmail.com> (cherry picked from commit 50ee56e)
Collaborator
🛠 PR Checks Summary🔴 Must not contain the "don't merge" label Manual Checks (for Reviewers):
Read More🤖 This bot helps streamline PR reviews by verifying automated checks and providing guidance for contributors and reviewers. ✅ Automated Checks (for Contributors):🔴 Must not contain the "don't merge" label ☑️ Contributor Actions:
☑️ Reviewer Actions:
📚 Resources:Debug
|
Codecov Report❌ Patch coverage is 📢 Thoughts on this report? Let us know! |
aeddi
changed the title
aeddi
changed the title
chain/gnoland1)
nemanjantic
added
a/gnops
nemanjantic
requested review from
D4ryl00,
davd-gzl,
gfanton,
jefft0,
jeronimoalbi,
julienrbrt,
ltzmaxwell,
notJoon,
omarsy,
sw360cab and
tbruyelle
github-project-automation
Bot
moved this from Triage
to Done
in 🧙♂️Gno.land development
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
20 participants
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Not meant to be merged. This PR is opened against
chain/gnoland1purely so reviewers can use GitHub's compare view — it gives a single click-through diff for every rc branch in the series and an easy "Files changed" tab. The actual work lands by cherry-picking from the rc branches into their respective upstream PRs (or via a coordinated hardfork fromchain/gnoland1directly). Don't press the green button.A stacked branch series that carries the gnoland1 → test-13 hardfork work. Each rc is a strict superset of the previous one — branch it off, add a focused delta, keep the older rcs intact so bisecting and reviewing never lose history. Open this PR against the highest rc (
chain/test13-rc6) unless you're reviewing one specific layer in isolation.A design/rationale writeup for the launch itself lives in the ADR:
gno.land/adr/pr5589_test13_hardfork_launch.md.Branch graph
Each row links to the branch tree and shows the cumulative diff against the immediate predecessor (what this rc adds).
chain/test13-base
Delta vs
chain/gnoland1: branches fromchain/gnoland1(the gnoland1 betanet genesis + its supporting core fixes, already merged upstream). Imports PR #5486 — the hf-glue testbed that builds a hardfork genesis.json from a source chain + its historical txs — plus the minimum number of small fixes needed to make #5486 run cleanly against the gnoland1 base. Nothing test13-specific yet: nor/sys/validators/v3, notest-13chain id, no post-fork migrations beyond what #5486 ships.hf_topup_balanceescape hatch for replay invariant failures$LOGvar inreport-replay.shmarkdown headerreport-replay.sh missing_importsreplay-log.shatgnogenesis fork testchain/test13-rc1
Delta vs
chain/test13-base: first end-to-end hardfork attempt. Imports the list of gno-core fixes curated on the launch-prep hackmd — these are the gnovm / tm2 / gnoland bugs we deemed worth carrying into the test-13 genesis (each referenced by its upstream PR number). Adds keyless genesis (so a genesis can be built without holding apriv_validator_key.json), the T1-rotation migration, and a chunked tx-archive fetch for flaky RPCs. Boots a single-validator test-13 locally;add-validatorvia v2 lands on-chain but the valset never moves — the problem rc2 + rc3 chase down.Delta vs
chain/gnoland1(cumulative)hf-glue specific
VALIDATOR_ADDR/VALIDATOR_PUBKEYmake upvia init dependencyinit-node.shidempotentgno-core fixes imported from the launch hackmd
doOpReffor correct interface pointer types (fix(gnovm): use static type in doOpRef for correct interface pointer types #5474)realm.SentCoins()(feat(stdlibs): implementrealm.SentCoins()#5039)IsDangerousURLto angle-bracket autolinks (fix(gnoweb): apply IsDangerousURL to angle-bracket autolinks #5515)r/sys/clarealm (feat: improve rendering of r/sys/cla realm #5331)p/nt/ownabledependency inr/sys/names(feat: remove p/nt/ownable dependency in r/sys/names #5244)AddressFromBytesandNewCoin(chore(std): Make non-panic versions of AddressFromBytes and NewCoin #5329)AssertOriginCallorigin check (fix(gnovm): skip closure frames inAssertOriginCallorigin check #5407)AddPart(fix(tm2): reject block parts with mismatched proofs in AddPart #5479)ArrayValue.Copy(fix(gnovm): deep-copy array elements in ArrayValue.Copy #5445)PrepareNewValues(fix(gnovm): track block item allocations in PrepareNewValues #5436)chain/test13-rc2
Delta vs
chain/test13-rc1: cherry-picks PR #5485 (valset-via-params v3), scaffolds a test13-specific deployment dir with the new v3-targetedadd-validator.sh, and plumbs the v3 realm into the post-fork migration. v3 doesn't deploy cleanly becauser/sys/names' namespace check rejects an addpkg intor/sys/*; this rc wraps the addpkg with a disable/restore pair. Also adds multi-validator genesis support infixvalidatorandmigrate.shso gno-cluster-style multi-node launches can be built.Delta vs
chain/gnoland1(cumulative)--valset-listr/sys/validators/v3in post-fork migrationadd-validator.shtargetingr/sys/validators/v3chain/test13-rc3
Delta vs
chain/test13-rc2: fixes the two reasonsadd-validator.shstill didn't move the valset on rc2. First,NewValsetChangeExecutorevaluates itschangesFneagerly so the executor no longer closes over the caller's/e/<caller>/runrealm (otherwisedao.MustCreateProposalpanics on "cannot persist function from the private realm"). Second, bakes a new migration step 08 that setsvm:p:valset_realm_path = gno.land/r/sys/validators/v3so EndBlocker reads from v3's paths. After rc3 the full loop works end-to-end.Delta vs
chain/gnoland1(cumulative)vm:p:valset_realm_pathto v3 in migrationchangesFneagerly inNewValsetChangeExecutorchain/test13-rc4
Delta vs
chain/test13-rc3: four tier-1 audit/assertion tools. None of rc3's replay failures panic the chain (absorbed by--skip-failing-genesis-txs), but silent divergence could ruin a launch. These close that hole:Delta vs
chain/gnoland1(cumulative)audit-balancesdiffing per-signer ugnot source vs replay — surfaces the 13 accounts drained by post-mainnet storage-deposit semanticsstate-difftool comparing replay vs source-chain realms — renders each realm on both sides + diffsverify-txs-jsonlintegrity check vs source-chain RPC — assertstotal_txs+ spot-checks random heightsassert-migrationsscript verifying post-replay state — one positive check per migration step's intended effectchain/test13-rc5
Delta vs
chain/test13-rc4: tier-2 resilience work. Proves the genesis rebuilds deterministically, teaches gno to survive mid-write process kills without crash-looping, adds every remaining validator-ops primitive beyond add, and patches one missing corner in the rc4 assertion script.Delta vs
chain/gnoland1(cumulative)assert-migrations— v3 init doesn't seed the key, so fresh-boot state legitimately presents as emptyrm/change-power/batchgovDAO scripts — full valset-ops surface against v3; change-power uses an atomic remove+add to work around v3's lack of in-place updatemempackage.go:889nil-pointer crash on cluster nodes killed mid-replay; defensive nil-skip so a half-persisted store boots instead of crash-loopingverify-reproducibilitybuilding genesis twice, asserting SHA match — local proxy for the cross-machine attestation validators will run at launchrc5 also proved, live against a 2-node cluster: add-validator → change-power → rm-validator → batch-add → mixed remove+update → halt-height set via govDAO → BaseApp panics at
block_height + 1exactly as designed.chain/test13-rc6
Delta vs
chain/test13-rc5: tier-3 audits and a documentation fix for a debugging pitfall.Delta vs
chain/gnoland1(cumulative)verify-reproducibilityassumes a clean OUT dir — after chasing phantom nondeterminism, the root cause was always staleout/between runs; comment-only update so the next debugger doesn't waste the same cyclecompare-gas-modesA/B-testing strict vs source replay —gnogenesis fork testruns twice, once per mode, diffs failure counts. Surfaces that on rc6 genesissourcemode eliminates zero failures (the 2580 InsufficientFunds all fire atante.DeductFeesbefore the gas meter engages), so the posture is "stay on strict / empty"audit-realm-importsflagging dangling imports post-fork — scans every addpkg tx (historical + genesis-mode) for imports that no longer resolve against the current stdlib + examples tree. Filters test-only files. Result: 368 unique(realm, import)edges, 0 danglingHow to review
add-validator.shfinally works, review rc2 + rc3 together.