feat: add --timing flag for phase-level performance breakdown (#2928)

* feat: add --timing flag for phase-level performance breakdown

Add PhaseTimer utility for instrumenting Slither phases:
- New --timing CLI flag to enable timing collection
- Instruments parse_contracts, analyze_contracts, convert_to_slithir,
  compute_dependency, compute_storage_layout, and detector runs
- Outputs sorted timing report showing total/count/avg per phase

Co-Authored-By: Claude Opus 4.5 <[email protected]>

* feat: add timing infrastructure and fix benchmark compatibility

Add phase-level timing support:
- New --timing flag for slither CLI to capture execution phases
- Timing data exported via --json-types timing
- Benchmark script updated to display phase breakdowns

Replace aave-v3-core with safe-smart-account in benchmarks:
- aave-v3-core requires Node 14/16/18, incompatible with Node 22+
- safe-smart-account uses modern Node.js (22.14.0 per .nvmrc)

Co-Authored-By: Claude Opus 4.5 <[email protected]>

---------

Co-authored-by: Claude Opus 4.5 <[email protected]>

scripts/benchmark_perf.py

@@ -24,6 +24,9 @@ Usage:
     # Compare against baseline and post to PR
     python scripts/benchmark_perf.py --slither-path /path/to/worktree \\
         --baseline /tmp/baseline.json --pr 123
+
+    # Include phase-level timing breakdown
+    python scripts/benchmark_perf.py --no-branch-compare --timing
 """
 
 import argparse
@@ -32,9 +35,11 @@ import json
 import os
 import subprocess
 import sys
+import tempfile
 import time
 from contextlib import contextmanager
 from pathlib import Path
+from typing import NamedTuple
 
 BENCH_DIR = Path("/tmp/slither-bench")
 LOCK_FILE = Path("/tmp/slither-bench.lock")
@@ -53,13 +58,13 @@ PROJECTS = {
         "url": "https://github.com/compound-finance/comet",
         "setup": ["yarn install"],
     },
-    "aave-v3-core": {
-        "url": "https://github.com/aave/aave-v3-core",
-        "setup": ["npm install"],
+    "safe-smart-account": {
+        "url": "https://github.com/safe-global/safe-smart-account",
+        "setup": ["npm install", "npx hardhat compile"],
     },
     "lido-dao": {
         "url": "https://github.com/lidofinance/lido-dao",
-        "setup": ["yarn install"],
+        "setup": ["yarn install", "forge build --build-info"],
     },
     "optimism": {
         "url": "https://github.com/ethereum-optimism/optimism",
@@ -123,20 +128,59 @@ def setup_projects(project_names: list[str]) -> None:
             subprocess.run(cmd, shell=True, cwd=work_dir, check=True)
 
 
-def run_slither(project_path: str, slither_cwd: Path | None = None) -> float:
-    """Run slither and return elapsed time.
+class BenchmarkResult(NamedTuple):
+    """Result from a single slither run."""
+
+    elapsed: float
+    timing: dict[str, dict] | None = None
+
+
+def run_slither(
+    project_path: str,
+    slither_cwd: Path | None = None,
+    capture_timing: bool = False,
+) -> BenchmarkResult:
+    """Run slither and return elapsed time and optional timing data.
 
     Args:
         project_path: Path to the project to analyze.
         slither_cwd: Directory to run slither from (for worktree support).
+        capture_timing: If True, capture phase-level timing data.
+
+    Returns:
+        BenchmarkResult with elapsed time and optional timing breakdown.
     """
-    start = time.perf_counter()
-    subprocess.run(
-        ["uv", "run", "slither", project_path],
-        capture_output=True,
-        cwd=slither_cwd,
-    )
-    return time.perf_counter() - start
+    timing_data = None
+    cmd = ["uv", "run", "slither", project_path]
+
+    if capture_timing:
+        # Use temp file to capture JSON with timing data
+        # Create path but don't keep the file (Slither won't overwrite existing files)
+        with tempfile.NamedTemporaryFile(suffix=".json", delete=True) as f:
+            json_path = f.name
+        # File is deleted when context exits, so Slither can create it fresh
+
+        cmd.extend(["--json", json_path, "--json-types", "timing", "--timing"])
+
+        start = time.perf_counter()
+        subprocess.run(cmd, capture_output=True, cwd=slither_cwd)
+        elapsed = time.perf_counter() - start
+
+        # Extract timing data from JSON
+        try:
+            with open(json_path) as f:
+                data = json.load(f)
+                timing_data = data.get("results", {}).get("timing")
+        except (json.JSONDecodeError, FileNotFoundError):
+            pass
+        finally:
+            Path(json_path).unlink(missing_ok=True)
+    else:
+        start = time.perf_counter()
+        subprocess.run(cmd, capture_output=True, cwd=slither_cwd)
+        elapsed = time.perf_counter() - start
+
+    return BenchmarkResult(elapsed=elapsed, timing=timing_data)
 
 
 def get_project_path(name: str) -> str:
@@ -148,17 +192,29 @@ def get_project_path(name: str) -> str:
     return str(base)
 
 
+class ProjectResult(NamedTuple):
+    """Result for a single project benchmark."""
+
+    elapsed: float
+    timing: dict[str, dict] | None = None
+
+
 def benchmark_current(
     project_names: list[str],
     quiet: bool = False,
     slither_cwd: Path | None = None,
-) -> dict[str, float]:
+    capture_timing: bool = False,
+) -> dict[str, ProjectResult]:
     """Benchmark current checkout against specified projects.
 
     Args:
         project_names: List of project names to benchmark.
         quiet: Suppress output if True.
         slither_cwd: Directory to run slither from (for worktree support).
+        capture_timing: If True, capture phase-level timing data.
+
+    Returns:
+        Dictionary mapping project name to ProjectResult.
     """
     results = {}
     for name in project_names:
@@ -168,11 +224,13 @@ def benchmark_current(
                 print(f"  {name}: SKIPPED (not found, run --setup first)")
             continue
 
-        times = [run_slither(path, slither_cwd) for _ in range(ITERATIONS)]
-        avg = sum(times) / len(times)
-        results[name] = round(avg, 1)
+        runs = [run_slither(path, slither_cwd, capture_timing) for _ in range(ITERATIONS)]
+        avg_elapsed = sum(r.elapsed for r in runs) / len(runs)
+        # Use timing from last run (they should be similar)
+        timing = runs[-1].timing if capture_timing else None
+        results[name] = ProjectResult(elapsed=round(avg_elapsed, 1), timing=timing)
         if not quiet:
-            print(f"  {name}: {avg:.1f}s")
+            print(f"  {name}: {avg_elapsed:.1f}s")
 
     return results
 
@@ -193,14 +251,39 @@ def benchmark_branches(project_names: list[str], branches: list[str], quiet: boo
     return results
 
 
+def summarize_timing(timing: dict[str, dict]) -> dict[str, float]:
+    """Summarize timing data into core phases and detector totals.
+
+    Args:
+        timing: Raw timing data from slither.
+
+    Returns:
+        Dictionary with summarized timing (core phases + detector_total).
+    """
+    core_phases = {}
+    detector_total = 0.0
+
+    for phase, data in timing.items():
+        total_sec = data.get("total_sec", 0)
+        if phase.startswith("detector:"):
+            detector_total += total_sec
+        else:
+            core_phases[phase] = total_sec
+
+    return {**core_phases, "detectors_total": round(detector_total, 3)}
+
+
 def format_results_markdown(
-    results: dict[str, float], baseline: dict[str, float] | None = None
+    results: dict[str, ProjectResult],
+    baseline: dict | None = None,
+    show_timing: bool = False,
 ) -> str:
     """Format benchmark results as markdown table.
 
     Args:
-        results: Benchmark results {project_name: time_seconds}.
+        results: Benchmark results {project_name: ProjectResult}.
         baseline: Optional baseline results for comparison.
+        show_timing: If True, include phase timing breakdown.
 
     Returns:
         Markdown-formatted table string.
@@ -211,17 +294,34 @@ def format_results_markdown(
         "| Project | Time | vs Baseline |",
         "|---------|------|-------------|",
     ]
-    for proj, time_s in results.items():
+    for proj, result in results.items():
+        time_s = result.elapsed
         if baseline and proj in baseline:
-            base_time = baseline[proj]
+            base_data = baseline[proj]
+            base_time = (
+                base_data.get("elapsed", base_data) if isinstance(base_data, dict) else base_data
+            )
             delta = ((base_time - time_s) / base_time) * 100 if base_time else 0
             lines.append(f"| {proj} | {time_s}s | {delta:+.1f}% |")
         else:
             lines.append(f"| {proj} | {time_s}s | - |")
+
+    if show_timing:
+        lines.append("")
+        lines.append("### Phase Breakdown")
+        lines.append("")
+        for proj, result in results.items():
+            if result.timing:
+                summary = summarize_timing(result.timing)
+                lines.append(f"**{proj}:**")
+                for phase, secs in sorted(summary.items(), key=lambda x: -x[1]):
+                    lines.append(f"- {phase}: {secs:.2f}s")
+                lines.append("")
+
     return "\n".join(lines)
 
 
-def load_baseline(baseline_path: Path) -> dict[str, float]:
+def load_baseline(baseline_path: Path) -> dict:
     """Load baseline results from JSON file.
 
     Args:
@@ -235,16 +335,20 @@ def load_baseline(baseline_path: Path) -> dict[str, float]:
 
 
 def print_pr_comment_command(
-    pr_number: int, results: dict[str, float], baseline: dict[str, float] | None = None
+    pr_number: int,
+    results: dict[str, ProjectResult],
+    baseline: dict | None = None,
+    show_timing: bool = False,
 ) -> None:
     """Print the gh command to post results as PR comment (no auto-posting).
 
     Args:
         pr_number: GitHub PR number.
-        results: Benchmark results {project_name: time_seconds}.
+        results: Benchmark results {project_name: ProjectResult}.
         baseline: Optional baseline results for comparison.
+        show_timing: If True, include phase timing in output.
     """
-    body = format_results_markdown(results, baseline)
+    body = format_results_markdown(results, baseline, show_timing)
     # Print markdown preview
     print("=== PR Comment Preview ===")
     print(body)
@@ -262,16 +366,43 @@ def print_comparison(results: dict) -> None:
 
     for branch, data in results.items():
         if branch == "master":
-            print(f"{branch}: {data}")
+            formatted = {k: f"{v.elapsed}s" for k, v in data.items()}
+            print(f"{branch}: {formatted}")
         else:
             changes = {}
-            for proj, t in data.items():
-                base = baseline.get(proj, t)
-                pct = ((base - t) / base) * 100 if base else 0
-                changes[proj] = f"{t}s ({pct:+.1f}%)"
+            for proj, result in data.items():
+                base_result = baseline.get(proj)
+                base_time = base_result.elapsed if base_result else result.elapsed
+                pct = ((base_time - result.elapsed) / base_time) * 100 if base_time else 0
+                changes[proj] = f"{result.elapsed}s ({pct:+.1f}%)"
             print(f"{branch}: {changes}")
 
 
+def print_timing_breakdown(results: dict[str, ProjectResult]) -> None:
+    """Print detailed timing breakdown for each project."""
+    print("\n=== Phase Timing Breakdown ===")
+    for proj, result in results.items():
+        if not result.timing:
+            continue
+        summary = summarize_timing(result.timing)
+        total = sum(summary.values())
+        print(f"\n{proj} ({result.elapsed}s total):")
+        for phase, secs in sorted(summary.items(), key=lambda x: -x[1]):
+            pct = (secs / total * 100) if total else 0
+            print(f"  {phase}: {secs:.2f}s ({pct:.0f}%)")
+
+
+def results_to_json(results: dict[str, ProjectResult]) -> dict:
+    """Convert ProjectResult dict to JSON-serializable format."""
+    return {
+        proj: {
+            "elapsed": result.elapsed,
+            "timing": summarize_timing(result.timing) if result.timing else None,
+        }
+        for proj, result in results.items()
+    }
+
+
 def main():
     parser = argparse.ArgumentParser(
         description="Benchmark Slither performance against smart contract projects",
@@ -324,6 +455,11 @@ def main():
         action="store_true",
         help="Skip lock acquisition (for isolated environments)",
     )
+    parser.add_argument(
+        "--timing",
+        action="store_true",
+        help="Capture and display phase-level timing breakdown",
+    )
 
     args = parser.parse_args()
 
@@ -373,7 +509,12 @@ def main():
         if args.no_branch_compare or args.slither_path:
             if not args.json:
                 print("=== Benchmark Results ===")
-            return benchmark_current(project_names, quiet=args.json, slither_cwd=slither_cwd)
+            return benchmark_current(
+                project_names,
+                quiet=args.json,
+                slither_cwd=slither_cwd,
+                capture_timing=args.timing,
+            )
         return benchmark_branches(project_names, branches, quiet=args.json)
 
     if args.no_lock:
@@ -384,15 +525,19 @@ def main():
 
     # Print PR comment command if requested (no auto-posting)
     if args.pr:
-        print_pr_comment_command(args.pr, results, baseline)
+        print_pr_comment_command(args.pr, results, baseline, show_timing=args.timing)
         return  # Don't print JSON when using --pr (output already includes preview)
 
     # Output results
     if args.json:
-        print(json.dumps(results, indent=2))
+        json_output = results_to_json(results)
+        print(json.dumps(json_output, indent=2))
     elif not args.no_branch_compare and not args.slither_path:
         print_comparison(results)
-        print("\n" + json.dumps(results, indent=2))
+        json_output = {branch: results_to_json(data) for branch, data in results.items()}
+        print("\n" + json.dumps(json_output, indent=2))
+    elif args.timing:
+        print_timing_breakdown(results)
 
 
 if __name__ == "__main__":

slither/__main__.py

@@ -649,6 +649,13 @@ def parse_args(
         default=False,
     )
 
+    parser.add_argument(
+        "--timing",
+        help="Print phase-level timing breakdown",
+        action="store_true",
+        default=False,
+    )
+
     # Disable the throw/catch on partial analyses
     parser.add_argument(
         "--disallow-partial", help=argparse.SUPPRESS, action="store_true", default=False
@@ -777,6 +784,14 @@ def main_impl(
         cp = cProfile.Profile()
         cp.enable()
 
+    # Enable phase timing if requested
+    if args.timing:
+        from slither.utils.timing import PhaseTimer
+
+        timer = PhaseTimer.get()
+        timer.reset()  # Clear accumulated data from previous runs
+        timer.enabled = True
+
     # Set colorization option
     set_colorization_enabled(False if args.disable_color else sys.stdout.isatty())
 
@@ -894,6 +909,12 @@ def main_impl(
                 _, printers = get_detectors_and_printers()
                 json_results["list-printers"] = output_printers_json(printers)
 
+            # Add timing data to JSON if requested
+            if "timing" in args.json_types and args.timing:
+                from slither.utils.timing import PhaseTimer
+
+                json_results["timing"] = PhaseTimer.get().report()
+
         # Output our results to markdown if we wish to compile a checklist.
         if args.checklist:
             output_results_to_markdown(
@@ -945,6 +966,13 @@ def main_impl(
         stats = pstats.Stats(cp).sort_stats("cumtime")
         stats.print_stats()
 
+    if args.timing:
+        from slither.utils.timing import PhaseTimer
+
+        # Skip text output when JSON goes to stdout to avoid corruption
+        if not outputting_json_stdout:
+            print("\n" + PhaseTimer.get().report_text())
+
     fail_on = FailOnLevel(args.fail_on)
     if fail_on == FailOnLevel.HIGH:
         fail_on_detection = any(result["impact"] == "High" for result in results_detectors)

slither/slither.py

@@ -197,9 +197,14 @@ class Slither(SlitherCore):
         self._init_parsing_and_analyses(kwargs.get("skip_analyze", False))
 
     def _init_parsing_and_analyses(self, skip_analyze: bool) -> None:
+        from slither.utils.timing import PhaseTimer
+
+        timer = PhaseTimer.get()
+
         for parser in self._parsers:
             try:
-                parser.parse_contracts()
+                with timer.phase("parse_contracts"):
+                    parser.parse_contracts()
             except Exception as e:
                 if self.no_fail:
                     continue
@@ -209,7 +214,8 @@ class Slither(SlitherCore):
         if not skip_analyze:
             for parser in self._parsers:
                 try:
-                    parser.analyze_contracts()
+                    with timer.phase("analyze_contracts"):
+                        parser.analyze_contracts()
                 except Exception as e:
                     if self.no_fail:
                         continue
@@ -282,9 +288,15 @@ class Slither(SlitherCore):
         """
         :return: List of registered detectors results.
         """
+        from slither.utils.timing import PhaseTimer
+
+        timer = PhaseTimer.get()
 
         self.load_previous_results()
-        results = [d.detect() for d in self._detectors]
+        results = []
+        for d in self._detectors:
+            with timer.phase(f"detector:{d.ARGUMENT}"):
+                results.append(d.detect())
 
         self.write_results_to_hide()
         return results

slither/solc_parsing/slither_compilation_unit_solc.py

@@ -582,12 +582,23 @@ class SlitherCompilationUnitSolc(CallerContextExpression):
         self._parsed = True
 
     def analyze_contracts(self) -> None:
+        from slither.utils.timing import PhaseTimer
+
         if not self._parsed:
             raise SlitherException("Parse the contract before running analyses")
-        self._convert_to_slithir()
+
+        timer = PhaseTimer.get()
+
+        with timer.phase("convert_to_slithir"):
+            self._convert_to_slithir()
+
         if not self._compilation_unit.core.skip_data_dependency:
-            compute_dependency(self._compilation_unit)
-        self._compilation_unit.compute_storage_layout()
+            with timer.phase("compute_dependency"):
+                compute_dependency(self._compilation_unit)
+
+        with timer.phase("compute_storage_layout"):
+            self._compilation_unit.compute_storage_layout()
+
         self._analyzed = True
 
     def _analyze_all_enums(self, contracts_to_be_analyzed: list[ContractSolc]) -> None:

slither/utils/command_line.py

@@ -25,6 +25,7 @@ JSON_OUTPUT_TYPES = [
     "printers",
     "list-detectors",
     "list-printers",
+    "timing",
 ]
 
 

slither/utils/timing.py

@@ -0,0 +1,70 @@
+"""Fine-grained timing instrumentation for Slither phases."""
+
+import json
+import threading
+import time
+from collections import defaultdict
+from collections.abc import Generator
+from contextlib import contextmanager
+
+
+class PhaseTimer:
+    """Singleton for collecting timing data across Slither phases."""
+
+    _instance: "PhaseTimer | None" = None
+    _lock: threading.Lock = threading.Lock()
+
+    def __init__(self) -> None:
+        self.timings: dict[str, list[float]] = defaultdict(list)
+        self.enabled = False
+
+    @classmethod
+    def get(cls) -> "PhaseTimer":
+        if cls._instance is None:
+            with cls._lock:
+                if cls._instance is None:
+                    cls._instance = PhaseTimer()
+        return cls._instance
+
+    @contextmanager
+    def phase(self, name: str) -> Generator[None, None, None]:
+        """Context manager to time a named phase."""
+        if not self.enabled:
+            yield
+            return
+        start = time.perf_counter()
+        try:
+            yield
+        finally:
+            self.timings[name].append(time.perf_counter() - start)
+
+    def report(self) -> dict[str, dict]:
+        """Return timing statistics sorted by total time."""
+        return {
+            name: {
+                "count": len(times),
+                "total_sec": round(sum(times), 3),
+                "avg_sec": round(sum(times) / len(times), 3) if times else 0,
+            }
+            for name, times in sorted(self.timings.items(), key=lambda x: sum(x[1]), reverse=True)
+        }
+
+    def report_json(self) -> str:
+        return json.dumps(self.report(), indent=2)
+
+    def report_text(self) -> str:
+        """Return human-readable timing report."""
+        lines = ["Phase Timing Report:"]
+        for name, stats in self.report().items():
+            total = stats["total_sec"]
+            count = stats["count"]
+            avg = stats["avg_sec"]
+            if count == 1:
+                lines.append(f"  {name}: {total:.3f}s")
+            else:
+                lines.append(f"  {name}: {total:.3f}s ({count} calls, avg {avg:.3f}s)")
+        return "\n".join(lines)
+
+    def reset(self) -> None:
+        """Clear all timing data."""
+        self.timings.clear()

tests/unit/utils/test_timing.py

@@ -0,0 +1,193 @@
+"""Unit tests for slither.utils.timing module."""
+
+import json
+import threading
+import time
+
+import pytest
+
+from slither.utils.timing import PhaseTimer
+
+
+@pytest.fixture(autouse=True)
+def reset_singleton():
+    """Reset the PhaseTimer singleton before and after each test."""
+    PhaseTimer._instance = None
+    yield
+    PhaseTimer._instance = None
+
+
+def test_phase_timer_disabled_by_default():
+    """Timer should be disabled by default."""
+    timer = PhaseTimer.get()
+    assert timer.enabled is False
+
+
+def test_phase_timer_no_collection_when_disabled():
+    """Timer should not collect timing data when disabled."""
+    timer = PhaseTimer.get()
+    assert timer.enabled is False
+
+    with timer.phase("test_phase"):
+        time.sleep(0.01)
+
+    assert len(timer.timings) == 0
+
+
+def test_phase_timer_collects_when_enabled():
+    """Timer should collect timing data when enabled."""
+    timer = PhaseTimer.get()
+    timer.enabled = True
+
+    with timer.phase("test_phase"):
+        time.sleep(0.01)
+
+    assert "test_phase" in timer.timings
+    assert len(timer.timings["test_phase"]) == 1
+    assert timer.timings["test_phase"][0] >= 0.01
+
+
+def test_phase_timer_multiple_calls():
+    """Timer should accumulate multiple calls to the same phase."""
+    timer = PhaseTimer.get()
+    timer.enabled = True
+
+    for _ in range(3):
+        with timer.phase("repeated_phase"):
+            time.sleep(0.001)
+
+    assert "repeated_phase" in timer.timings
+    assert len(timer.timings["repeated_phase"]) == 3
+
+
+def test_phase_timer_report_structure():
+    """Report should have correct structure with count, total_sec, and avg_sec."""
+    timer = PhaseTimer.get()
+    timer.enabled = True
+
+    with timer.phase("phase_a"):
+        time.sleep(0.01)
+
+    with timer.phase("phase_b"):
+        time.sleep(0.02)
+
+    report = timer.report()
+
+    assert "phase_a" in report
+    assert "phase_b" in report
+
+    for phase_name in ["phase_a", "phase_b"]:
+        assert "count" in report[phase_name]
+        assert "total_sec" in report[phase_name]
+        assert "avg_sec" in report[phase_name]
+        assert report[phase_name]["count"] == 1
+
+
+def test_phase_timer_report_text_format():
+    """Report text should be human-readable with proper formatting."""
+    timer = PhaseTimer.get()
+    timer.enabled = True
+
+    with timer.phase("my_phase"):
+        time.sleep(0.01)
+
+    text = timer.report_text()
+
+    assert "Phase Timing Report:" in text
+    assert "my_phase" in text
+    assert "s" in text  # Time suffix
+
+
+def test_phase_timer_reset():
+    """Reset should clear all timing data."""
+    timer = PhaseTimer.get()
+    timer.enabled = True
+
+    with timer.phase("phase_to_clear"):
+        time.sleep(0.001)
+
+    assert len(timer.timings) == 1
+
+    timer.reset()
+
+    assert len(timer.timings) == 0
+
+
+def test_phase_timer_singleton():
+    """PhaseTimer.get() should always return the same instance."""
+    timer1 = PhaseTimer.get()
+    timer2 = PhaseTimer.get()
+
+    assert timer1 is timer2
+
+
+def test_phase_timer_singleton_thread_safe():
+    """Singleton should be thread-safe under concurrent access."""
+    instances = []
+    errors = []
+
+    def get_instance():
+        try:
+            instance = PhaseTimer.get()
+            instances.append(instance)
+        except Exception as e:
+            errors.append(e)
+
+    threads = [threading.Thread(target=get_instance) for _ in range(10)]
+    for t in threads:
+        t.start()
+    for t in threads:
+        t.join()
+
+    assert len(errors) == 0
+    assert len(instances) == 10
+    # All instances should be the same object
+    assert all(inst is instances[0] for inst in instances)
+
+
+def test_phase_timer_report_json():
+    """report_json() should return valid JSON string."""
+    timer = PhaseTimer.get()
+    timer.enabled = True
+
+    with timer.phase("json_test"):
+        time.sleep(0.001)
+
+    json_str = timer.report_json()
+    parsed = json.loads(json_str)
+
+    assert "json_test" in parsed
+    assert parsed["json_test"]["count"] == 1
+
+
+def test_phase_timer_report_sorted_by_total_time():
+    """Report should be sorted by total time in descending order."""
+    timer = PhaseTimer.get()
+    timer.enabled = True
+
+    with timer.phase("short"):
+        time.sleep(0.001)
+
+    with timer.phase("long"):
+        time.sleep(0.02)
+
+    report = timer.report()
+    phases = list(report.keys())
+
+    # "long" should come before "short" since it has more total time
+    assert phases.index("long") < phases.index("short")
+
+
+def test_phase_timer_multiple_phases_text():
+    """Report text should show call count and average for multiple calls."""
+    timer = PhaseTimer.get()
+    timer.enabled = True
+
+    for _ in range(3):
+        with timer.phase("multi_call"):
+            time.sleep(0.001)
+
+    text = timer.report_text()
+
+    assert "3 calls" in text
+    assert "avg" in text