- Add code_to_optimize/**/package-lock.json to .gitignore
- Re-enable Dependabot version updates with limit of 5 PRs per ecosystem
- Keep code_to_optimize/ ignore comment in dependabot.yml
pytest-cov's trace function conflicts with the Tracer class under test,
causing it to self-disable in CI. Linux also reports ~1% lower coverage
than macOS due to platform-specific branches.
Assigns per-directory code ownership to current org members based on
full commit history analysis, so PRs automatically request reviews from
the right people.
- Add pytest-cov to dev dependencies
- Add .coveragerc with branch coverage, 60% floor (current baseline),
and source/omit configuration
- Add coverage CI job (ubuntu/py3.13) that runs pytest with --cov,
enforces the floor, and uploads coverage.xml as an artifact
- Wire coverage into the required-checks-passed gate
Closes#2080
- Add PR template with required linked issue/discussion section
- Add check-linked-issue CI job that validates PR body contains a
reference (#123, Closes/Fixes/Relates, GitHub URL, or CF-# ticket)
- Wire into required-checks-passed gate so it blocks merge
- Update CONTRIBUTING.md with the policy and motivation
Closes#522
Covers the two audiences the issue calls out:
1. People contributing changes back to Codeflash - development
environment setup with uv, the single-command verification via
uv run prek, test runner invocation, code-style pointers to
.claude/rules/code-style.md, and the branch / commit / PR
conventions from .claude/rules/git.md and CLAUDE.md.
2. People using Codeflash in editable mode from a source checkout
to optimize their own projects, including the install commands
for uv and pip, when editable mode is appropriate vs installing
the PyPI package, and a pointer to the README Quick Start for
the codeflash init flow.
Keep the combined JFR + tracing agent single JVM invocation from main while
preserving the fix's intent: raise when trace-db was not created, warn when
exit code is non-zero but trace-db exists. Integration tests rewritten to
match the combined-invocation semantics.
Gradle evaluates all project configurations during the configuration
phase, even when only one module is targeted. Multi-module projects with
diverse toolchain requirements (e.g., OpenRewrite's rewrite-gradle needs
JDK 8) fail when an unrelated module's toolchain isn't available.
Adds --configure-on-demand to all 8 Gradle command construction sites
so Gradle only configures projects needed for the requested task.
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
_run_java_with_graceful_timeout() discarded the subprocess exit code in
both the no-timeout and timeout paths. If Maven/Gradle failed (compilation
error, OOM, etc.), the tracer silently continued with missing/stale data.
Now returns the exit code. Stage 1 (JFR profiling) warns on failure but
continues. Stage 2 (argument capture) raises RuntimeError since trace
data is essential for replay test generation.
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
Keep only repeatString which reliably produces 284% improvement.
Drop computeSum (marginal 16%), filterEvens and instanceMethod (no
optimization found). Reduces tracer E2E from ~1h27m to ~21m.
Combine JFR profiling and argument capture agent into one
JAVA_TOOL_OPTIONS string, running the target program once instead of
twice. JFR and javaagent are orthogonal JVM features that coexist
without conflict. Keeps build_jfr_env/build_agent_env for standalone
use.
Make TOTAL_LOOPING_TIME configurable via CODEFLASH_LOOPING_TIME env var
(defaults to 10s). Set to 5s in Java E2E CI jobs to cut verification
time per candidate. Also cache the codeflash-runtime JAR keyed on
source hash to skip mvn install when unchanged.
Add .resolve() to TemporaryDirectory path to expand Windows 8.3 short
paths (e.g. RUNNER~1) to canonical long form, fixing test_pickle_patcher
failures on Windows CI. Also add missing return type annotations and
noqa suppressions for benchmark test file.
Move codeflash's own benchmarks to .codeflash/benchmarks/. Add
auto-discovery of .codeflash/benchmarks/ in codeflash compare and
benchmark mode -- when benchmarks-root is not explicitly configured,
the CLI checks for .codeflash/benchmarks/ before erroring.
Backwards compatible: users with existing benchmarks-root config
are unaffected. Docs continue to show tests/benchmarks as the
example path.
JavaSupport.ensure_runtime_environment() was never called during the
optimization flow, so _language_version stayed None and the backend
received language_version=null. The LLM had no Java version constraint,
causing it to generate Java 16+ APIs (e.g. Stream.toList()) for Java 11
projects.
Set PYTHONHASHSEED=0 in test subprocess environments so original and
candidate runs use identical hash behavior, eliminating a source of
non-deterministic return-value comparisons.
Also upgrade diff logging from debug to info level with actual types
and repr values for DID_PASS, RETURN_VALUE, and STDOUT diffs.
The jdk.ExecutionSample#period=1ms syntax in -XX:StartFlightRecording
is only supported on JDK 13+. On JDK 11 (CI), it causes
"Failure when starting JFR on_create_vm_2" and no JFR file is created.
The settings=profile preset still provides 10ms CPU sampling.
