Merge pull request #21 from OpenArchive/feat/veilid-0.5.1-upgrade

Veilid dependency upgrade

Author: tripledoublev

.cargo/config-android.toml

@@ -8,3 +8,7 @@ ar = "llvm-ar"
 [target.armv7-linux-androideabi]
 linker = "armv7a-linux-androideabi34-clang"
 ar = "llvm-ar"
+
+[target.x86_64-linux-android]
+linker = "x86_64-linux-android34-clang"
+ar = "llvm-ar"

.config/nextest.toml

@@ -0,0 +1,40 @@
+# Nextest configuration for save-rust
+# Veilid/DHT tests can be slow and flaky due to P2P peer discovery and network setup.
+
+[profile.default]
+# Set timeout to 10 minutes per test (Veilid network operations can be slow)
+# Terminate after 10 periods of 60s.
+slow-timeout = { period = "60s", terminate-after = 10 }
+
+# Retry flaky tests with exponential backoff to give Veilid time to connect
+retries = { backoff = "exponential", count = 3, delay = "5s", max-delay = "30s" }
+
+# Run tests serially to avoid Veilid network conflicts
+test-threads = 1
+
+# P2P tests are flaky in CI, but very high retry counts can stretch failures to >1h.
+# Keep retries moderate so truly broken runs fail fast.
+[[profile.default.overrides]]
+filter = 'test(test_replicate_group) | test(test_join_group)'
+retries = { backoff = "exponential", count = 4, delay = "8s", max-delay = "45s", jitter = true }
+
+[[profile.default.overrides]]
+filter = 'test(test_refresh_joined_group)'
+retries = { backoff = "exponential", count = 2, delay = "8s", max-delay = "30s", jitter = true }
+slow-timeout = { period = "60s", terminate-after = 5 }
+
+# Faster CI profile used for PR runs.
+[profile.ci-virtual]
+slow-timeout = { period = "30s", terminate-after = 4 }
+retries = { backoff = "exponential", count = 1, delay = "3s", max-delay = "10s", jitter = true }
+test-threads = 1
+
+[[profile.ci-virtual.overrides]]
+filter = 'test(test_replicate_group) | test(test_join_group)'
+retries = { backoff = "exponential", count = 2, delay = "5s", max-delay = "20s", jitter = true }
+slow-timeout = { period = "45s", terminate-after = 4 }
+
+[[profile.ci-virtual.overrides]]
+filter = 'test(test_refresh_joined_group)'
+retries = { backoff = "exponential", count = 1, delay = "5s", max-delay = "15s", jitter = true }
+slow-timeout = { period = "45s", terminate-after = 4 }

.github/workflows/lint_and_test.yml

@@ -7,8 +7,10 @@ on:
       - main
 
 jobs:
-  lint:
+  lint_and_test:
     runs-on: ubuntu-latest
+    timeout-minutes: 75
+
     steps:
       - name: Set up Rust toolchain
         uses: hecrj/setup-rust-action@v2
@@ -24,37 +26,18 @@ jobs:
       - name: Run Clippy
         run: cargo clippy --all-targets --all-features -- -D warnings
 
-  test:
-    needs: lint
-    runs-on: ubuntu-latest
-    timeout-minutes: 15
-    strategy:
-      matrix:
-        test_name: [
-          "tests::basic_test",
-          "tests::test_health_endpoint",
-          "tests::test_join_group",
-          "tests::test_refresh_empty_group",
-          "tests::test_refresh_group_with_file",
-          "tests::test_refresh_group_with_single_repo",
-          # Skipping: flaky P2P peer discovery (see issue #TODO)
-          # "tests::test_refresh_joined_group",
-          "tests::test_refresh_nonexistent_group",
-          # Skipping: flaky P2P peer discovery (see issue #TODO)
-          # "tests::test_replicate_group",
-          "tests::test_upload_list_delete"
-        ]
-
-    steps:
-      - name: Set up Rust toolchain
-        uses: hecrj/setup-rust-action@v2
-        with:
-          rust-version: stable
-      
-      - name: Check out the code
-        uses: actions/checkout@v4
+      - name: Install cargo-nextest
+        run: cargo install cargo-nextest --locked
 
-      - name: Run individual test
+      - name: Run tests with retries
         env:
           RUST_MIN_STACK: 8388608
-        run: cargo test --verbose -- ${{ matrix.test_name }} --test-threads=1 --nocapture
+          SAVE_VEILID_LOCAL_TEST_MODE: "1"
+        # Retries configured in .config/nextest.toml (nextest default path)
+        # --no-fail-fast: run all tests in this PR subset even when some fail.
+        # Exclude long P2P network convergence tests from PR CI; run them in dedicated/nightly workflows.
+        run: >
+          cargo nextest run
+          --profile ci-virtual
+          --no-fail-fast
+          -E 'not (test(test_join_group) | test(test_replicate_group) | test(test_refresh_joined_group) | test(test_idempotent_create_repo_after_join))'

.gitignore

@@ -2,3 +2,8 @@
 /target
 platform-build
 private-*
+test_*.log
+.cursor/
+save-data/
+test-output.txt
+test_refresh_joined_group_output.txt

API.md

@@ -414,11 +414,14 @@ Request Body: Binary file content
 Response:
 ```json
 {
-    "name": "string",                // File name
-    "updated_collection_hash": "string" // Hash of the updated collection
+    "name": "string",                 // File name
+    "updated_collection_hash": "string", // Hash of the updated collection
+    "file_hash": "string"             // Hash of the uploaded file
 }
 ```
 
+`file_hash` is the Veilid content hash stored in the DHT. Clients can use it to verify local uploads, deduplicate media, or trigger replication via the refresh endpoints if the hash is missing locally.
+
 Error Response (400 Bad Request):
 ```json
 {

CHANGELOG.md

@@ -0,0 +1,6 @@
+# Changelog
+
+## 2026-02-03
+
+- Upgrade to `veilid-core` v0.5.1, applying the temporary `fix-underflow` patch while the upstream bugfix is under review.
+- Adopt `cargo-nextest` with retries in CI to stabilize Veilid/DHT-related tests.