fix --binary-path for tracer source repo checkouts

[p-datadog]

Summary

Right now, --binary-path works great for artifact-based tracers (Java jars, .NET tarballs, PHP packages) where the contents are flat files. But for libraries that support installing from a source repo checkout, it silently does the wrong thing — and that feels like it could be a real pain point for local development.

Here's what happens today if you try the natural invocation:

./build.sh ruby --binary-path ~/code/dd-trace-rb

build.sh runs cp -r ~/code/dd-trace-rb/* binaries/, spreading repo contents flat across binaries/. The Ruby install script checks for /binaries/dd-trace-rb (a named subdirectory), doesn't find it, and silently falls back to installing from rubygems.org instead of your local code. No warning, no error — just not testing what you think you're testing.

This affects 7 libraries that expect a named subdirectory inside binaries/:

Library	Expected dir	Silent fallback target
Ruby	dd-trace-rb	rubygems.org
Node.js	dd-trace-js	npm registry
Go	dd-trace-go	Go module proxy
Python	dd-trace-py	PyPI
Rust	dd-trace-rs	crates.io / GitHub
C++	dd-trace-cpp	GitHub clone
C++ Kong	dd-trace-cpp	GitHub clone

What this PR does

• Adds a get_expected_repo_dir() function that maps each library to its expected source repo directory name (e.g., ruby → dd-trace-rb)
• When --binary-path is used and the basename of the path matches the expected repo dir, copies into binaries/<repo-dir>/ using rsync with symlink dereferencing (--copy-links) and .git exclusion — with a check that rsync is available
• When the basename doesn't match but the library expects a repo dir, prints a warning to stderr so the silent fallback is no longer silent
• When the basename doesn't match, preserves the existing flat cp -r behavior unchanged — so artifact-based workflows and staging directories keep working exactly as before

Why rsync for the source repo path?

Two reasons: symlink dereferencing and .git exclusion. Local repos often contain symlinks (git worktrees, bundler binstubs, monorepo package links) that would be broken inside the Docker container. And .git directories are large and unnecessary for the build.

This is the same approach that utils/scripts/watch.sh already uses for syncing local repos into binaries/.

How it works in practice

# These now all work as you'd expect:
./build.sh ruby   --binary-path ~/code/dd-trace-rb
./build.sh nodejs --binary-path ~/code/dd-trace-js
./build.sh golang --binary-path ~/code/dd-trace-go
./build.sh python --binary-path ~/code/dd-trace-py
./build.sh rust   --binary-path ~/code/dd-trace-rs

# Non-standard names get a warning instead of silent fallback:
./build.sh ruby --binary-path ~/code/dtr
# Warning: --binary-path basename 'dtr' does not match expected 'dd-trace-rb'.
#   Doing flat copy. If this is a source checkout, rename the directory to 'dd-trace-rb'.

# Existing artifact workflows are unchanged (no warning, plain cp -r):
./build.sh java   --binary-path ~/builds/java-artifacts   # flat jars
./build.sh dotnet --binary-path ~/builds/dotnet-output     # flat .so/.tar.gz

The documented workflow of cloning directly into binaries/ also continues to work — this just makes --binary-path do the right thing too.

Test plan

• Verify --binary-path ~/code/dd-trace-rb correctly creates binaries/dd-trace-rb/ and installs from local source (Ruby)
• Verify --binary-path ~/builds/java-artifacts still does flat copy for artifact-based tracers (Java)
• Verify staging directory workflow still works: --binary-path /tmp/staging where staging contains dd-trace-rb/
• Verify non-standard directory names produce a warning
• Verify symlinks in source repos are dereferenced correctly
• Verify .git directory is excluded from the copy

Generated with Claude Code

[github-actions]

CODEOWNERS have been resolved as:

utils/build/build.sh                                                    @DataDog/system-tests-core

[datadog-official]

⚠️ Tests

✨ Fix all issues with BitsAI or with Cursor

⚠️ Warnings

🧪 4 Tests failed

tests.integrations.test_dsm.Test_Dsm_Manual_Checkpoint_Inter_Process.test_dsm_manual_checkpoint_inter_process[spring-boot] from system_tests_suite (Datadog) (Fix with Cursor)

ValueError: Checkpoint has not been found, please have a look in logs

self = <tests.integrations.test_dsm.Test_Dsm_Manual_Checkpoint_Inter_Process object at 0x7f2bf534d640>

    def test_dsm_manual_checkpoint_inter_process(self):
        assert self.produce_threaded.status_code == 200
        assert self.produce_threaded.text == "ok"
        assert "dd-pathway-ctx-base64" in self.produce_threaded.headers
    
        assert self.consume_threaded.status_code == 200
...

tests.integrations.test_dsm.Test_Dsm_Manual_Checkpoint_Inter_Process.test_dsm_manual_checkpoint_inter_process[spring-boot] from system_tests_suite (Datadog) (Fix with Cursor)

ValueError: Checkpoint has not been found, please have a look in logs

self = <tests.integrations.test_dsm.Test_Dsm_Manual_Checkpoint_Inter_Process object at 0x7fbb348ec770>

    def test_dsm_manual_checkpoint_inter_process(self):
        assert self.produce_threaded.status_code == 200
        assert self.produce_threaded.text == "ok"
        assert "dd-pathway-ctx-base64" in self.produce_threaded.headers
    
        assert self.consume_threaded.status_code == 200
...

tests.parametric.test_dynamic_configuration.TestDynamicConfigV1.test_trace_sampling_rate_override_env[library_env0, parametric-golang] from system_tests_suite (Datadog) (Fix with Cursor)

AssertionError: No RemoteConfig apply status found, got requests []

self = <tests.parametric.test_dynamic_configuration.TestDynamicConfigV1 object at 0x7fd8034e3890>
library_env = {'DD_ENV': 'test_env', 'DD_INTERNAL_TELEMETRY_V2_ENABLED': '1', 'DD_REMOTE_CONFIG_POLL_INTERVAL_SECONDS': '0.2', 'DD_SERVICE': 'test_service', ...}
test_agent = <utils.docker_fixtures._test_agent.TestAgentAPI object at 0x7fd7ceb30230>
test_library = <utils.docker_fixtures._test_clients._test_client_parametric.ParametricTestClientApi object at 0x7fd7ce8e0290>

    @parametrize(
        "library_env",
        [{"DD_TRACE_SAMPLE_RATE": r, **DEFAULT_ENVVARS} for r in ["0.1", "1.0"]],
...

tests.parametric.test_library_tracestats.Test_Library_Tracestats.test_distinct_aggregationkeys_TS003[agent_env0-library_env0, parametric-golang] from system_tests_suite (Datadog) (Fix with Cursor)

AssertionError: There should be one bucket containing the stats
assert 2 == 1
 +  where 2 = len([{'Duration': 10000000000, 'Start': 1774038100000000000, 'Stats': [{'Duration': 2935247, 'ErrorSummary': store: {}}, m...key:0, offset:0, zero_count: 0.0, count: 0.0, sum: 0.0, min: inf, max: -inf, 'Errors': 0, 'GRPCStatusCode': '', ...}]}])

self = <tests.parametric.test_library_tracestats.Test_Library_Tracestats object at 0x7f31cab9e6f0>
test_agent = <utils.docker_fixtures._test_agent.TestAgentAPI object at 0x7f3195f42030>
test_library = <utils.docker_fixtures._test_clients._test_client_parametric.ParametricTestClientApi object at 0x7f3195c3b320>

    @enable_tracestats()
    @enable_agent_version()
...

View all

ℹ️ Info

No other issues found (see more)

❄️ No new flaky tests detected

_{This comment will be updated automatically if new data arrives.
🔗 Commit SHA: e27c0fb | Docs | Datadog PR Page | Was this helpful? React with 👍/👎 or give us feedback!}