build: mention @claude in github to get work done#552
Conversation
Installs protoc v25.2, clippy, rustfmt, and system dependencies (clang, cmake, libsqlite3-dev, libzmq3-dev) to match CI workflows. Only runs in Claude Code remote environments. https://claude.ai/code/session_01SdEvmPYuJkJoAVJejB8Tz9
The tenderdash-proto build.rs downloads proto sources from GitHub via ureq/rustls during compilation, which fails in sandboxed environments where TLS certificate verification doesn't work (InvalidCertificate). The session-start hook now: - Pre-downloads the Tenderdash v1.5.3 zip archive via curl (which uses the system CA bundle and works in sandboxed environments) - Extracts sources to /tmp/tenderdash-v1.5.3 (TENDERDASH_DIR) - Caches the zip at /tmp/tenderdash-cache/ (CARGO_TARGET_DIR for the proto-compiler's archive lookup) - Sets both env vars via ~/.cargo/config.toml [env] so the build script finds the cached archive and skips the network download Validated: cargo build, cargo clippy, and cargo test all pass. https://claude.ai/code/session_01SdEvmPYuJkJoAVJejB8Tz9
|
Note Reviews pausedIt looks like this branch is under active development. To avoid overwhelming you with review comments due to an influx of new commits, CodeRabbit has automatically paused this review. You can configure this behavior by changing the Use the following commands to manage reviews:
Use the checkboxes below for quick actions:
📝 WalkthroughWalkthroughAdds Claude Code integration infrastructure including a session-start Bash script for remote environment setup (protoc, Rust, system packages, pre-downloaded dependencies), configuration file defining hooks, and GitHub Actions workflow enabling Claude Code execution on repository interactions. Changes
Sequence DiagramsequenceDiagram
actor Developer
participant GitHub as GitHub
participant Workflow as GitHub Actions
participant ClaudeAction as Claude Code Action
participant Remote as Remote Environment
participant Script as session-start.sh
Developer->>GitHub: Comment with `@claude`
GitHub->>Workflow: Trigger workflow
Workflow->>ClaudeAction: Execute Claude Code Action
ClaudeAction->>Remote: Connect to remote environment
Remote->>Script: Execute session-start.sh
Script->>Script: Install protoc (if needed)
Script->>Script: Install Rust components
Script->>Script: Install system packages
Script->>Script: Pre-download Tenderdash sources
Script->>Script: Configure Cargo environment
Script->>Remote: Return ready environment
Remote->>ClaudeAction: Environment initialized
ClaudeAction->>Developer: Ready for code operations
Estimated code review effort🎯 3 (Moderate) | ⏱️ ~25 minutes Poem
🚥 Pre-merge checks | ✅ 2 | ❌ 1❌ Failed checks (1 inconclusive)
✅ Passed checks (2 passed)
✏️ Tip: You can configure your own custom pre-merge checks in the settings. ✨ Finishing touches🧪 Generate unit tests (beta)
Thanks for using CodeRabbit! It's free for OSS, and your support helps us grow. If you like it, consider giving us a shout-out. Comment |
|
@claude check if this PR builds |
There was a problem hiding this comment.
Actionable comments posted: 1
🤖 Fix all issues with AI agents
In @.claude/hooks/session-start.sh:
- Around line 103-111: The script appends a second [env] table which produces
invalid TOML if CARGO_CONFIG_FILE already contains an [env] section; update
session-start.sh to detect whether an [env] table exists (searching for a line
matching "^\[env\]") and if present, ensure TENDERDASH_DIR and CARGO_TARGET_DIR
keys are added/updated beneath that existing [env] block (instead of appending a
new [env] header); if no [env] exists, create the [env] block as currently done.
Use CARGO_CONFIG_FILE, TENDERDASH_DIR and TENDERDASH_CACHE_DIR as the target
symbols to find and insert/update the key=value lines atomically (or employ a
TOML-aware tool) so the file remains valid TOML.
🧹 Nitpick comments (1)
.claude/hooks/session-start.sh (1)
60-60: Hardcoded Tenderdash version requires manual sync withbuild.rs.The comment on Line 59 notes this must match
DEFAULT_VERSIONinrs-tenderdash-abci/proto/build.rs. If that version is bumped without updating this script, the pre-download becomes useless and builds will attempt (and potentially fail) network fetches again. Consider extracting this to a shared constant or adding a CI check that validates the versions match.
- Add "matcher": "startup" to SessionStart hook so it only runs on new sessions, not on resume/clear/compact (deps are persistent) - Quote $CLAUDE_PROJECT_DIR for paths with spaces - Add explicit 120s timeout for the hook - Fix cargo config append to handle pre-existing [env] sections without creating duplicate TOML table headers https://claude.ai/code/session_01SdEvmPYuJkJoAVJejB8Tz9
- Replace archived actions-rs/toolchain@v1 with dtolnay/rust-toolchain - Replace archived actions-rs/clippy-check@v1 with direct cargo clippy - Replace archived actions-rs/cargo@v1 with direct cargo commands - Fix push branch trigger from `main` to `master` (the actual release branch) https://claude.ai/code/session_01SdEvmPYuJkJoAVJejB8Tz9
|
Hardcoded Tenderdash version requires manual sync with build.rs. The comment on Line 59 notes this must match @claude Implement a solution that will detect required version of tenderdash automatically. |
Adds GitHub Actions integration for Claude Code with OAuth authentication. The workflow triggers on @claude mentions in issues, PR comments, and reviews. Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
|
@claude check if you can build this project |
There was a problem hiding this comment.
Actionable comments posted: 1
🤖 Fix all issues with AI agents
In @.github/workflows/clippy.yml:
- Around line 56-60: The "Install Rust toolchain" step currently uses
dtolnay/rust-toolchain@master with toolchain: "1.92", which diverges from
tests.yml's dtolnay/rust-toolchain@stable; update the clippy workflow to match
the tests workflow strategy by either switching the "uses" to
dtolnay/rust-toolchain@stable (and remove the explicit "toolchain" pin) or pin
both workflows to the same explicit toolchain string (e.g., change toolchain in
clippy.yml and tests.yml so both use "1.92"), editing the Install Rust toolchain
step (uses and/or toolchain fields) so both workflows run the identical Rust
version.
🧹 Nitpick comments (4)
.github/workflows/clippy.yml (1)
67-74:PROTOCenv var is step-scoped — won't be visible during buildThe
env: PROTOCon lines 73–74 is scoped to the "Install protoc" step only. If anybuild.rs(e.g., protobuf code generation) checks thePROTOCenvironment variable rather than searchingPATH, it won't find it during the clippy or fmt steps. Sinceprotocis installed to/usr/local/bin(which is onPATH), this likely works in practice, but it would be cleaner to either move thePROTOCenv to the job level or remove it entirely.Suggested fix: promote to job-level env
env: CARGO_TERM_COLOR: always + PROTOC: /usr/local/bin/protocand remove the step-level
envblock on lines 73–74..github/workflows/tests.yml (1)
56-57: Consider pinning the Rust toolchain version to match clippy.ymlUsing
@stablewithout a version pin means the tests workflow will automatically pick up new stable releases. While this catches forward-compatibility issues early, it also means CI can break unpredictably on a new Rust release day. The clippy workflow pins"1.92"— aligning both would give more deterministic builds.Suggested change
- name: Install Rust toolchain - uses: dtolnay/rust-toolchain@stable + uses: dtolnay/rust-toolchain@master + with: + toolchain: "1.92".github/workflows/claude.yml (2)
21-26: Broad write permissions — ensure this is intentional.
contents: writegrants Claude the ability to push commits directly. Combined withpull-requests: writeandissues: write, this gives the action significant repository access. This is expected for the Claude Code action workflow, but worth confirming the team is comfortable with this scope — especially since any GitHub user who can comment@claudeon a public issue/PR could trigger it.Consider whether restricting triggers to specific users/teams (e.g., via an additional
ifcheck ongithub.actor) would be appropriate for your threat model.
37-37: User-specific secret creates a bus-factor risk.
CLAUDE_CODE_OAUTH_TOKEN_LKLIMEKis tied to a single developer. If that person leaves or their token is rotated, this workflow silently breaks. Consider using a team/org-level secret (e.g.,CLAUDE_CODE_OAUTH_TOKEN) or a machine account token instead.
…ithub_token Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
There was a problem hiding this comment.
Actionable comments posted: 1
🤖 Fix all issues with AI agents
In @.github/workflows/claude.yml:
- Around line 21-26: Remove the unnecessary id-token: write permission from the
workflow permissions block (the id-token key) since this job uses direct
Anthropic API authentication via anthropic_api_key rather than cloud provider
authentication; update the permissions list to omit id-token while leaving other
permissions (contents, pull-requests, issues, actions) unchanged to reduce the
workflow's security surface.
|
@claude in session-start.sh, update protoc to the most recent version |
1 similar comment
|
@claude in session-start.sh, update protoc to the most recent version |
There was a problem hiding this comment.
Caution
Some comments are outside the diff and can’t be posted inline due to platform limitations.
⚠️ Outside diff range comments (1)
CLAUDE.md (1)
17-17:⚠️ Potential issue | 🟡 MinorRemove nightly requirement from formatting instructions.
Lines 17 and 36 document conflicting formatting commands: line 17 specifies
cargo fmt --all, while line 36 specifiescargo +nightly fmt. Sincerust-toolchain.tomlpins the project to stable Rust 1.92 and no nightly-specific rustfmt configuration exists, the+nightlyrequirement should be removed.Change line 36 from
cargo +nightly fmttocargo fmt --allto match line 17.
🧹 Nitpick comments (1)
CLAUDE.md (1)
9-9: Clarify which conventional commit standard to follow.The guideline mentions "conventional commit naming rules" but doesn't specify which standard. Consider referencing the Conventional Commits specification (e.g.,
feat:,fix:,docs:,chore:, etc.) or provide examples to make this more actionable for contributors.📝 Suggested clarification
-- PR and commits should follow conventional commit naming rules. +- PR and commit messages should follow the [Conventional Commits](https://www.conventionalcommits.org/) specification (e.g., `feat:`, `fix:`, `docs:`, `chore:`).
Cherry-pick from v1.0-dev to enable @claude mentions in PRs. Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
…561) * fix: return error instead of panicking on corrupted database blobs (#560) - Replace `unreachable!()` in `get_scheduled_votes` with warning log and default to false for unexpected `executed` column values - Change `QualifiedIdentity::from_bytes()` to return `Result` instead of panicking via `.expect()` - Propagate deserialization errors as `rusqlite::Error` in all 6 callers so corrupted database is surfaced to the user rather than silently ignored or crashing the app - Add `CorruptedBlobError` newtype in database module to eliminate repeated `FromSqlConversionFailure` boilerplate Closes #560 Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com> * chore: impl thiserror * chore: fmt * build: add Claude Code GitHub workflow and settings (#552) Cherry-pick from v1.0-dev to enable @claude mentions in PRs. Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com> * chore: fix after merge * fix: skip corrupted identity blobs in get_wallets instead of aborting Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com> * chore: fail on corrupted identity * doc: document error handling in the db --------- Co-authored-by: Claude Opus 4.6 <noreply@anthropic.com>
* fix: skip best chain lock polling in SPV mode (#567) * Initial plan * Skip chain lock refresh in SPV mode Co-authored-by: lklimek <842586+lklimek@users.noreply.github.com> * Document SPV guard intent Co-authored-by: lklimek <842586+lklimek@users.noreply.github.com> --------- Co-authored-by: copilot-swe-agent[bot] <198982749+Copilot@users.noreply.github.com> Co-authored-by: lklimek <842586+lklimek@users.noreply.github.com> * fix: compute relative timestamps from actual data (#581) * fix: compute relative timestamps from actual data Replace hardcoded display strings like "Received: 1 day ago" with real relative timestamps computed from document createdAt/updatedAt fields using chrono::Utc::now(). Fixes #579 * style: fix import ordering per cargo fmt * refactor: extract format_relative_time to shared dashpay module Deduplicate the identical format_relative_time function that existed in both contact_requests.rs and send_payment.rs. Move it to the dashpay mod.rs as a pub(crate) function and import from both files. --------- Co-authored-by: PastaClaw <thepastaclaw@users.noreply.github.com> * fix: update platform for DIP-18 HRP and improve SPV sync progress (#575) * fix: update dashpay/platform to d6f4eb9 for DIP-18 HRP fix The previous platform revision used incorrect bech32m HRP prefixes (evo/tevo) for Platform addresses. The updated commit uses the correct DIP-0018 prefixes (dash/tdash). Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com> * refactor(spv): use dash-spv SyncProgress API directly Replace the intermediate SyncProgress/DetailedSyncProgress/SyncStage translation layer with direct use of dash_spv::sync::SyncProgress. This eliminates ~120 lines of bridge code in spawn_progress_watcher() and determine_sync_stage(), and lets the UI query per-manager progress (headers, filter_headers, filters, masternodes, blocks) via the upstream API. Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com> * fix(spv): improve sync progress accuracy and robustness - Log headers() error before discarding for easier debugging - Restore download-window optimization for headers progress on checkpoint-resumed syncs (progress starts near 0% not 83%) - Replace catch-all _ => 0.0 with explicit SyncState variant matches so new variants produce compile errors - Fix filters progress to use current_height/target_height instead of downloaded/target_height (session count vs absolute height mismatch) - Restore peer count in sync status text - Add "Querying peer heights" label and diffs_processed to masternode status for more informative sync messages Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com> * fix(spv): use height-based blocks progress to prevent bar from jumping The blocks progress bar was bouncing backward because it used processed/requested session counters whose denominator grows as filters discover more matching blocks. Switch to last_processed block height relative to headers target_height, which only increases. Display "current / target" heights instead of percentage on the blocks bar. Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com> * fix(spv): improve progress bar resilience across resume and network switch - Add windowed progress tracking for filters (consistent with headers/filter_headers) - Reset blocks_stage_start on Error state so recovery gets a fresh window - Track spv_progress_network to detect network changes and rebuild progress state from the new network's sync_progress instead of resetting to zero - Eliminate redundant clone in progress watcher (move instead of clone twice) - Consolidate all progress state reset logic into rebuild_spv_progress_state() Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com> --------- Co-authored-by: Claude Opus 4.6 <noreply@anthropic.com> * build: add git and cargo permissions to Claude Code workflow (#565) * build: add git and cargo permissions to Claude Code workflow Allow Claude to run git fetch/merge/checkout/rebase/push and cargo build/test/clippy/fmt commands. Switch model to opus. Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com> * fix: use claude_args for model and allowed tools The `model` and `allowed_tools` inputs are not declared in claude-code-action@v1 and are silently ignored. Move them to `claude_args` with --model and --allowedTools flags. Also fix deprecated colon syntax (`:*`) to space syntax (` *`). Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com> * fix: use single quotes to prevent glob expansion in allowed tools Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com> * build: sandbox cargo commands and add git permissions for Claude - Add safe-cargo.sh wrapper that strips CI secrets before running cargo - Use --allowedTools for git and safe-cargo, --disallowedTools for raw cargo - Document safe-cargo usage in CLAUDE.md Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com> * fix: switch safe-cargo.sh from denylist to allowlist approach Use `env -i` (start with empty environment, explicitly pass only what cargo needs) instead of `env -u` (strip known secrets). This is more robust against future secrets being added to the workflow. Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com> * build: deny Claude from editing CI scripts and workflows Prevent Claude from modifying .github/scripts/ and .github/workflows/ to ensure the safe-cargo wrapper cannot be tampered with. Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com> * fix: conditionally pass optional env vars in safe-cargo.sh Empty PROTOC="" caused prost build scripts to fail with "protoc not found". Now optional vars (PROTOC, CC, CXX, etc.) are only passed when set and non-empty. Tested: build, test, fmt all pass through the wrapper. Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com> * rabbit feedback * build: move safe-cargo.sh to scripts/ and allow +nightly fmt Move safe-cargo.sh from .github/scripts/ to top-level scripts/ for better discoverability. Add detailed comment explaining why the wrapper exists (prevent CI secret exfiltration via build scripts). Update all references in claude.yml, CLAUDE.md, and permission settings. Add `+nightly fmt` to allowedTools so Claude can follow CLAUDE.md formatting instructions in CI. Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com> --------- Co-authored-by: Claude Opus 4.6 <noreply@anthropic.com> * fix: handle errors instead of panicking on corrupted database blobs (#561) * fix: return error instead of panicking on corrupted database blobs (#560) - Replace `unreachable!()` in `get_scheduled_votes` with warning log and default to false for unexpected `executed` column values - Change `QualifiedIdentity::from_bytes()` to return `Result` instead of panicking via `.expect()` - Propagate deserialization errors as `rusqlite::Error` in all 6 callers so corrupted database is surfaced to the user rather than silently ignored or crashing the app - Add `CorruptedBlobError` newtype in database module to eliminate repeated `FromSqlConversionFailure` boilerplate Closes #560 Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com> * chore: impl thiserror * chore: fmt * build: add Claude Code GitHub workflow and settings (#552) Cherry-pick from v1.0-dev to enable @claude mentions in PRs. Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com> * chore: fix after merge * fix: skip corrupted identity blobs in get_wallets instead of aborting Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com> * chore: fail on corrupted identity * doc: document error handling in the db --------- Co-authored-by: Claude Opus 4.6 <noreply@anthropic.com> * feat: track DAPI endpoint availability in connection status (#533) * fix: connection status not updated * chore: rabbit feedback * Initial plan * Track DAPI connection status and display in tooltips and connection info - Add dapi_total_endpoints and dapi_available fields to ConnectionStatus - Factor DAPI availability into overall_connected status (RED when no endpoints available) - Query SDK AddressList during periodic refresh for endpoint counts and availability - Display DAPI status in connection indicator tooltip - Display DAPI status in network chooser Connection Status card (all modes) - Add dapi_status_label() helper for consistent status text formatting Co-authored-by: lklimek <842586+lklimek@users.noreply.github.com> * Address review feedback: store available endpoint count and extract DRY helper - Changed dapi_available from AtomicBool to AtomicU16 (dapi_available_endpoints) to store the count of available endpoints instead of just a boolean - Display format now shows "Available ({available}/{total} endpoints)" - Extracted repeated DAPI status rendering into add_dapi_status_label() helper function in network_chooser_screen.rs to eliminate code duplication Co-authored-by: lklimek <842586+lklimek@users.noreply.github.com> * Fix rustfmt formatting issues in connection_status.rs - Reorder atomic imports (AtomicU8 before AtomicU16) per rustfmt - Wrap long .store() call to respect line length - Wrap long format!() strings to respect line length Co-authored-by: lklimek <842586+lklimek@users.noreply.github.com> * Fix borrow checker error: extract DAPI status values before mutable self borrow The add_dapi_status_label helper was capturing a &ConnectionStatus reference (derived from self) in closures, which extended the immutable borrow past the self.render_spv_sync_progress() mutable borrow on line 594. Fix: change add_dapi_status_label to accept pre-computed owned values (dapi_total, dapi_available, dapi_label) instead of &ConnectionStatus, and extract those values early alongside other status fields. Co-authored-by: lklimek <842586+lklimek@users.noreply.github.com> * Switch Platform to c2c88e4 and use get_live_addresses() for accurate available count Updated dash-sdk dependency to Platform commit c2c88e4a988ce930 which adds AddressList::get_live_addresses() method. Replaced the workaround that used get_live_address().is_some() (which could only tell if at least one endpoint was live) with get_live_addresses().len() to get the exact count of available non-banned DAPI endpoints. Co-authored-by: lklimek <842586+lklimek@users.noreply.github.com> * chore: typo + network changes * chore: apply feedback * chore: rabbit review * chore: rabbit feedback * chore: rabbitting * Remove overall_connected_with method (deleted upstream in base branch) Co-authored-by: lklimek <842586+lklimek@users.noreply.github.com> * Merge v1.0-dev into copilot/update-dapi-connection-status Resolved modify/delete conflict on src/context.rs: removed the file since v1.0-dev refactored it into src/context/mod.rs and submodules, which already include our Arc<ConnectionStatus> changes. Co-authored-by: lklimek <842586+lklimek@users.noreply.github.com> * chore: fix platfom versioning issues * chore: add todo * chore: fmt --------- Co-authored-by: Lukasz Klimek <842586+lklimek@users.noreply.github.com> Co-authored-by: copilot-swe-agent[bot] <198982749+Copilot@users.noreply.github.com> * refactor(spv): eliminate separate thread/runtime, improve shutdown and UX (#577) * fix: update dashpay/platform to d6f4eb9 for DIP-18 HRP fix The previous platform revision used incorrect bech32m HRP prefixes (evo/tevo) for Platform addresses. The updated commit uses the correct DIP-0018 prefixes (dash/tdash). Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com> * refactor(spv): use dash-spv SyncProgress API directly Replace the intermediate SyncProgress/DetailedSyncProgress/SyncStage translation layer with direct use of dash_spv::sync::SyncProgress. This eliminates ~120 lines of bridge code in spawn_progress_watcher() and determine_sync_stage(), and lets the UI query per-manager progress (headers, filter_headers, filters, masternodes, blocks) via the upstream API. Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com> * fix(spv): improve sync progress accuracy and robustness - Log headers() error before discarding for easier debugging - Restore download-window optimization for headers progress on checkpoint-resumed syncs (progress starts near 0% not 83%) - Replace catch-all _ => 0.0 with explicit SyncState variant matches so new variants produce compile errors - Fix filters progress to use current_height/target_height instead of downloaded/target_height (session count vs absolute height mismatch) - Restore peer count in sync status text - Add "Querying peer heights" label and diffs_processed to masternode status for more informative sync messages Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com> * fix(spv): use height-based blocks progress to prevent bar from jumping The blocks progress bar was bouncing backward because it used processed/requested session counters whose denominator grows as filters discover more matching blocks. Switch to last_processed block height relative to headers target_height, which only increases. Display "current / target" heights instead of percentage on the blocks bar. Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com> * fix(spv): improve progress bar resilience across resume and network switch - Add windowed progress tracking for filters (consistent with headers/filter_headers) - Reset blocks_stage_start on Error state so recovery gets a fresh window - Track spv_progress_network to detect network changes and rebuild progress state from the new network's sync_progress instead of resetting to zero - Eliminate redundant clone in progress watcher (move instead of clone twice) - Consolidate all progress state reset logic into rebuild_spv_progress_state() Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com> * fix(spv): update ConnectionStatus immediately on connect/disconnect Propagate SPV status into ConnectionStatus right after start_spv() and stop_spv() so the UI reflects the change on the next frame instead of waiting for the next throttled trigger_refresh() cycle (2-10 seconds). Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com> * fix(spv): check cancellation token in listener tasks to prevent shutdown hang SPV finality and reconcile listeners blocked app shutdown for up to 10s because they never checked the global cancellation token. Add cancel branches to their tokio::select! loops so they exit immediately on shutdown. Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com> * chore: remove log spamming line * fix(spv): make listener handler calls cancellation-aware and add shutdown trace logging Wrap reconcile_spv_wallets(), handle_spv_finality_event(), and the debounce sleep inside tokio::select! with the cancellation token so shutdown can interrupt them even when blocked on locks held by the SPV sync thread. Add trace-level logging to TaskManager::shutdown() for per-task join timing to aid future shutdown diagnostics. Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com> * fix(spv): disable Disconnect button during stopping and poll status faster - Disable the Disconnect button while SPV status is Stopping to prevent double-clicks and provide visual feedback - Poll SPV status every 200ms during Stopping instead of the 10s connected interval so the Stopped transition is reflected within 1s - Reset the throttle timer in stop_spv() so fast polling starts immediately Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com> * refactor(tasks): add mandatory task names to spawn_sync for shutdown diagnostics Change spawn_sync() signature to require a &'static str name. The JoinSet now yields the task name on completion, letting shutdown() log which tasks finished and which ones timed out. Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com> * chore: log task names * refactor(spv): eliminate separate OS thread and tokio runtime Replace the dedicated SPV thread + 4-worker tokio runtime with a spawned task on the main 12-worker runtime via TaskManager::spawn_sync. This simplifies shutdown (SPV loop now tracked in unified JoinSet), removes cross-runtime complexity, and improves debuggability. Additional changes: - Fix: zeroize xprv_str after wallet import (security H-1) - Fix: sanitize devnet_name in build_spv_data_dir to prevent path traversal - Add 21 integration tests covering lifecycle, concurrency, deadlock detection, and live testnet sync Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com> * fix(spv): derive stop_token as child of global cancel for clean shutdown The spv_request_handler selected on stop_token, which was independent of the global cancellation token. During window-close shutdown only the global token was cancelled, leaving the request handler running and causing a ~5s hang until the TaskManager timeout aborted it. Fix by creating stop_token as a child_token() of the global cancel. This also simplifies run_spv_loop and run_sync_and_monitor by removing the redundant global_cancel parameter — a single stop_token now covers both explicit SpvManager::stop() and application-wide shutdown. Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com> * CLAUDE: minor changes about docs * fix(spv): bump rust-dashcore to d8bc066 for faster disconnect and add shutdown tracing Pin rust-dashcore patches to commit d8bc066 which includes: - sync manager task loop exits on network errors instead of logging indefinitely - sync coordinator signal_shutdown() cancels tasks before network disconnect - connection tasks race Peer::connect() against shutdown token Add debug tracing to SpvManager::stop() and run_sync_and_monitor() to measure client.stop() duration, and trace-level polling of SPV status in ConnectionStatus. Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com> * chore: update rust-dashcore rev * merge: resolve v1.0-dev conflicts for #577 (#585) * fix: compute relative timestamps from actual data (#581) * fix: compute relative timestamps from actual data Replace hardcoded display strings like "Received: 1 day ago" with real relative timestamps computed from document createdAt/updatedAt fields using chrono::Utc::now(). Fixes #579 * style: fix import ordering per cargo fmt * refactor: extract format_relative_time to shared dashpay module Deduplicate the identical format_relative_time function that existed in both contact_requests.rs and send_payment.rs. Move it to the dashpay mod.rs as a pub(crate) function and import from both files. --------- Co-authored-by: PastaClaw <thepastaclaw@users.noreply.github.com> * fix: update platform for DIP-18 HRP and improve SPV sync progress (#575) * fix: update dashpay/platform to d6f4eb9 for DIP-18 HRP fix The previous platform revision used incorrect bech32m HRP prefixes (evo/tevo) for Platform addresses. The updated commit uses the correct DIP-0018 prefixes (dash/tdash). Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com> * refactor(spv): use dash-spv SyncProgress API directly Replace the intermediate SyncProgress/DetailedSyncProgress/SyncStage translation layer with direct use of dash_spv::sync::SyncProgress. This eliminates ~120 lines of bridge code in spawn_progress_watcher() and determine_sync_stage(), and lets the UI query per-manager progress (headers, filter_headers, filters, masternodes, blocks) via the upstream API. Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com> * fix(spv): improve sync progress accuracy and robustness - Log headers() error before discarding for easier debugging - Restore download-window optimization for headers progress on checkpoint-resumed syncs (progress starts near 0% not 83%) - Replace catch-all _ => 0.0 with explicit SyncState variant matches so new variants produce compile errors - Fix filters progress to use current_height/target_height instead of downloaded/target_height (session count vs absolute height mismatch) - Restore peer count in sync status text - Add "Querying peer heights" label and diffs_processed to masternode status for more informative sync messages Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com> * fix(spv): use height-based blocks progress to prevent bar from jumping The blocks progress bar was bouncing backward because it used processed/requested session counters whose denominator grows as filters discover more matching blocks. Switch to last_processed block height relative to headers target_height, which only increases. Display "current / target" heights instead of percentage on the blocks bar. Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com> * fix(spv): improve progress bar resilience across resume and network switch - Add windowed progress tracking for filters (consistent with headers/filter_headers) - Reset blocks_stage_start on Error state so recovery gets a fresh window - Track spv_progress_network to detect network changes and rebuild progress state from the new network's sync_progress instead of resetting to zero - Eliminate redundant clone in progress watcher (move instead of clone twice) - Consolidate all progress state reset logic into rebuild_spv_progress_state() Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com> --------- Co-authored-by: Claude Opus 4.6 <noreply@anthropic.com> * build: add git and cargo permissions to Claude Code workflow (#565) * build: add git and cargo permissions to Claude Code workflow Allow Claude to run git fetch/merge/checkout/rebase/push and cargo build/test/clippy/fmt commands. Switch model to opus. Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com> * fix: use claude_args for model and allowed tools The `model` and `allowed_tools` inputs are not declared in claude-code-action@v1 and are silently ignored. Move them to `claude_args` with --model and --allowedTools flags. Also fix deprecated colon syntax (`:*`) to space syntax (` *`). Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com> * fix: use single quotes to prevent glob expansion in allowed tools Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com> * build: sandbox cargo commands and add git permissions for Claude - Add safe-cargo.sh wrapper that strips CI secrets before running cargo - Use --allowedTools for git and safe-cargo, --disallowedTools for raw cargo - Document safe-cargo usage in CLAUDE.md Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com> * fix: switch safe-cargo.sh from denylist to allowlist approach Use `env -i` (start with empty environment, explicitly pass only what cargo needs) instead of `env -u` (strip known secrets). This is more robust against future secrets being added to the workflow. Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com> * build: deny Claude from editing CI scripts and workflows Prevent Claude from modifying .github/scripts/ and .github/workflows/ to ensure the safe-cargo wrapper cannot be tampered with. Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com> * fix: conditionally pass optional env vars in safe-cargo.sh Empty PROTOC="" caused prost build scripts to fail with "protoc not found". Now optional vars (PROTOC, CC, CXX, etc.) are only passed when set and non-empty. Tested: build, test, fmt all pass through the wrapper. Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com> * rabbit feedback * build: move safe-cargo.sh to scripts/ and allow +nightly fmt Move safe-cargo.sh from .github/scripts/ to top-level scripts/ for better discoverability. Add detailed comment explaining why the wrapper exists (prevent CI secret exfiltration via build scripts). Update all references in claude.yml, CLAUDE.md, and permission settings. Add `+nightly fmt` to allowedTools so Claude can follow CLAUDE.md formatting instructions in CI. Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com> --------- Co-authored-by: Claude Opus 4.6 <noreply@anthropic.com> --------- Co-authored-by: PastaClaw <thepastaclaw@users.noreply.github.com> Co-authored-by: lklimek <842586+lklimek@users.noreply.github.com> Co-authored-by: Claude Opus 4.6 <noreply@anthropic.com> * chore: bump rust-dashcore * chore: imports --------- Co-authored-by: Claude Opus 4.6 <noreply@anthropic.com> Co-authored-by: Pasta Lil Claw <pasta+claw@dashboost.org> Co-authored-by: PastaClaw <thepastaclaw@users.noreply.github.com> Co-authored-by: PastaPastaPasta <6443210+PastaPastaPasta@users.noreply.github.com> * ci: cancel in-progress workflow runs on new push (#590) * ci: cancel in-progress workflow runs on new push Add concurrency groups to Tests and Clippy workflows so that previous runs are automatically cancelled when a new commit is pushed to the same branch or PR. Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com> * trigger ci --------- Co-authored-by: Claude Opus 4.6 <noreply@anthropic.com> * fix: use correct DashPay profile field "publicMessage" instead of "bio" (#582) The DashPay contract schema defines the field as "publicMessage", not "bio". This caused profile bios to never load in contact info. Co-authored-by: PastaClaw <thepastaclaw@users.noreply.github.com> --------- Co-authored-by: Copilot <198982749+Copilot@users.noreply.github.com> Co-authored-by: lklimek <842586+lklimek@users.noreply.github.com> Co-authored-by: PastaClaw <thepastaclaw@users.noreply.github.com> Co-authored-by: Claude Opus 4.6 <noreply@anthropic.com> Co-authored-by: PastaPastaPasta <6443210+PastaPastaPasta@users.noreply.github.com>
Installs Claude Code so that we can mention @claude during code review and give it some tasks.
Uses my (@lklimek) personal Claude subscription for now. Can be replaced with Claude API key in the future.
This pull request introduces automation and environment setup improvements for Claude Code integration, ensuring consistency between local development and CI environments. It adds a session start hook to install required dependencies and pre-download proto sources, configures Claude Code hooks, and introduces a GitHub Actions workflow for Claude Code. Additionally, it updates development guidelines to enforce conventional commit naming.
Claude Code integration and automation:
.github/workflows/claude.ymlto run Claude Code workflows in response to comments or PR reviews mentioning@claude, ensuring Claude Code automation is triggered appropriately..claude/settings.jsonto configure a session start hook for Claude Code environments, running the new setup script on startup.Development environment setup:
.claude/hooks/session-start.shscript to automate installation of required system packages, Rust components,protoc, and to pre-download Tenderdash proto sources, ensuring local environments match CI and builds do not fail due to missing dependencies or network issues.Development process guidelines:
CLAUDE.mdto require PRs and commits to follow conventional commit naming rules, promoting consistency in commit history.Installs protoc v25.2, clippy, rustfmt, and system dependencies
(clang, cmake, libsqlite3-dev, libzmq3-dev) to match CI workflows.
Only runs in Claude Code remote environments.
https://claude.ai/code/session_01SdEvmPYuJkJoAVJejB8Tz9
Summary by CodeRabbit
Chores
Documentation