Trigger CI

Author: gyuheon0h

.github/workflows/test.yml

@@ -269,7 +269,6 @@ jobs:
     needs: setup
     if: needs.setup.outputs.packages-count != '0'
     runs-on: ubuntu-latest
-    timeout-minutes: 720
     env:
       PACKAGES: ${{ needs.setup.outputs.packages }}
       CRASHTRACKER_FEATURE: ${{ needs.setup.outputs.crashtracker-feature }}
@@ -298,7 +297,7 @@ jobs:
           cache-bin: true # cache the ~/.cargo/bin directory
       - name: Build CentOS 7 Docker image
         run: docker build -t libdatadog-centos7 -f tools/docker/Dockerfile.centos .
-      - name: "cargo nextest run (centos7) x100"
+      - name: "cargo nextest run (centos7)"
         # Run docker as a user, not the default root
         # Mount and use the runner's toolchain as it's the same arch
         # exclude tracing integration tests since they require docker in docker to run a test-agent
@@ -311,20 +310,17 @@ jobs:
             # shellcheck disable=SC2086
             NEXTEST_CMD="cargo nextest run $PACKAGES $CRASHTRACKER_FEATURE --profile ci --no-tests=pass -E '!test(tracing_integration_tests::)'"
           fi
-          for i in $(seq 1 100); do
-            echo "=== Run $i/100 ==="
-            docker run --rm \
-              --user "$(id -u):$(id -g)" \
-              -v "${{ github.workspace }}:/workspace" \
-              -v "${CARGO_HOME:-$HOME/.cargo}:/usr/local/cargo" \
-              -v "${RUSTUP_HOME:-$HOME/.rustup}:/usr/local/rustup" \
-              -e CARGO_HOME=/usr/local/cargo \
-              -e RUSTUP_HOME=/usr/local/rustup \
-              -e PATH=/usr/local/cargo/bin:/usr/local/rustup/toolchains/stable-x86_64-unknown-linux-gnu/bin:/opt/rh/devtoolset-11/root/usr/bin:/usr/local/sbin:/usr/local/bin:/usr/sbin:/usr/bin:/sbin:/bin \
-              -w /workspace \
-              libdatadog-centos7 \
-              sh -c "$NEXTEST_CMD"
-          done
+          docker run --rm \
+            --user "$(id -u):$(id -g)" \
+            -v "${{ github.workspace }}:/workspace" \
+            -v "${CARGO_HOME:-$HOME/.cargo}:/usr/local/cargo" \
+            -v "${RUSTUP_HOME:-$HOME/.rustup}:/usr/local/rustup" \
+            -e CARGO_HOME=/usr/local/cargo \
+            -e RUSTUP_HOME=/usr/local/rustup \
+            -e PATH=/usr/local/cargo/bin:/usr/local/rustup/toolchains/stable-x86_64-unknown-linux-gnu/bin:/opt/rh/devtoolset-11/root/usr/bin:/usr/local/sbin:/usr/local/bin:/usr/sbin:/usr/bin:/sbin:/bin \
+            -w /workspace \
+            libdatadog-centos7 \
+            sh -c "$NEXTEST_CMD"
       - name: Add file attributes to JUnit XML
         if: success() || failure()
         run: cargo run --bin add_junit_file_attributes -- target/nextest/ci/junit.xml

libdd-crashtracker/src/receiver/ptrace_collector.rs

@@ -408,8 +408,12 @@ fn is_transient_ptrace_error(err: &PtraceError) -> bool {
 
 /// Attempt to capture a thread context, retrying once on transient failures.
 ///
-/// Uses a single per-thread deadline across attempts so that a slow thread
-/// cannot consume more than STOP_TIMEOUT_PER_THREAD total.
+/// Each attempt gets its own `STOP_TIMEOUT_PER_THREAD` budget (capped at the
+/// overall deadline) so that a retry after a timeout-induced failure actually
+/// has enough time to succeed. On older kernels (e.g. CentOS 7 / kernel 3.10)
+/// the first attempt can consume its entire budget waiting for registers to
+/// become readable; reusing that exhausted deadline would make the retry a
+/// no-op.
 ///
 /// A capture that succeeds but produces zero frames is also retried: on a
 /// running thread with a confirmed non-zero IP, empty frames indicates a
@@ -435,13 +439,15 @@ fn capture_with_retry(
         return None;
     }
 
-    let remaining = thread_deadline.saturating_duration_since(Instant::now());
-    if remaining <= RETRY_DELAY {
+    if Instant::now() + RETRY_DELAY >= overall_deadline {
         return None;
     }
     std::thread::sleep(RETRY_DELAY);
 
-    capture_thread_context(tid, resolve_frames, addr_space, thread_deadline).ok()
+    // Give the retry a fresh per-thread budget so it has a real chance of
+    // succeeding even if the first attempt consumed its entire deadline.
+    let retry_deadline = (Instant::now() + STOP_TIMEOUT_PER_THREAD).min(overall_deadline);
+    capture_thread_context(tid, resolve_frames, addr_space, retry_deadline).ok()
 }
 
 /// Stream thread contexts to a callback one at a time.