Syntax highlight and hyperlink crash functions + agent sections (#1139)

This PR adds syntax highlighting to the stack traces and turns file
paths into clickable links that point directly to the source code on
GitHub using the `get_project_repository` method in
[`oss_fuzz_checkout.py`](https://github.com/google/oss-fuzz-gen/blob/46693b387e7e32c38a2206692bf4440a5b86c072/experiment/oss_fuzz_checkout.py#L183).

## Preview


![image](https://github.com/user-attachments/assets/9dc75773-c53c-4b0c-a063-fe6d31a3aec8)

Author: myanvoos

experiment/builder_runner.py

@@ -1032,23 +1032,50 @@ def build_and_run_cloud(
       if blob.exists():
         blob.download_to_file(f)
 
+    code_coverage_report_dir = self.work_dirs.code_coverage_report(
+        generated_target_name)
+    report_prefix = f'{coverage_name}/report/linux/'
+    blobs = bucket.list_blobs(prefix=report_prefix)
+
+    for blob in blobs:
+      if blob.name.endswith('/'):
+        continue
+
+      # Get the relative path within report/linux/
+      relative_path = blob.name[len(report_prefix):]
+      local_path = os.path.join(code_coverage_report_dir, 'report', 'linux',
+                                relative_path)
+      os.makedirs(os.path.dirname(local_path), exist_ok=True)
+
+      blob.download_to_filename(local_path)
+
     run_result = RunResult(corpus_path=corpus_path,
                            coverage_report_path=coverage_path,
                            reproducer_path=reproducer_path,
                            log_path=run_log_path)
 
-    blob = bucket.blob(f'{coverage_name}/report/linux/summary.json')
-    if blob.exists():
-      # Download summary.json to our workdir.
-      cov_summary_folder = os.path.join(
-          self.work_dirs.code_coverage_report(generated_target_name),
-          'report/linux/')
-      os.makedirs(cov_summary_folder, exist_ok=True)
-      coverage_summary_file = os.path.join(cov_summary_folder, 'summary.json')
-      with open(coverage_summary_file, 'wb') as f:
-        blob.download_to_file(f)
-
-      # Load the coverage summary
+    #     blob = bucket.blob(f'{coverage_name}/report/linux/summary.json')
+    # if blob.exists():
+    #   # Download summary.json to our workdir.
+    #   cov_summary_folder = os.path.join(
+    #       self.work_dirs.code_coverage_report(generated_target_name),
+    #       'report/linux/')
+    #   os.makedirs(cov_summary_folder, exist_ok=True)
+    #   coverage_summary_file = os.path.join(cov_summary_folder, 'summary.json')
+    #   with open(coverage_summary_file, 'wb') as f:
+    #     blob.download_to_file(f)
+
+    # # Load the coverage summary
+    # with open(coverage_summary_file, 'r') as f:
+    #   run_result.coverage_summary = json.load(f)
+
+    # summary.json is already downloaded as part of the bulk download above
+    coverage_summary_file = os.path.join(
+        self.work_dirs.code_coverage_report(generated_target_name),
+        'report/linux/summary.json')
+
+    # Load the coverage summary if it exists
+    if os.path.exists(coverage_summary_file):
       with open(coverage_summary_file, 'r') as f:
         run_result.coverage_summary = json.load(f)
 

logger.py

@@ -67,15 +67,15 @@ def write_result(self,
     with open(os.path.join(trial_result_dir, FINAL_RESULT_JSON), 'w') as f:
       json.dump(result.to_dict() | {'finished': finished}, f)
 
-  def write_chat_history(self, result: Result) -> None:
+  def write_chat_history(self, result: Result, cycle_count: int) -> None:
     """Writes chat history."""
     # TODO(dongge): Find a proper way to write this.
     trial_result_dir = os.path.join(result.work_dirs.status,
                                     f'{result.trial:02d}')
     os.makedirs(trial_result_dir, exist_ok=True)
     chat_history_path = os.path.join(trial_result_dir, 'log.txt')
     chat_history = '\n'.join(
-        f'\n\n\n************************{agent_name}************************\n'
+        f'\n\n\n************************{agent_name} (Cycle {cycle_count})************************\n'
         f'{chat_history}\n'
         for agent_name, chat_history in result.chat_history.items())
     self.write_to_file(chat_history_path, chat_history)

pipeline.py

@@ -112,7 +112,8 @@ def _execute_one_cycle(self, result_history: list[Result],
     # and if it fails then we will return from this cycle and terminate
     # the pipeline.
     result_history.append(
-        self.writing_stage.execute(result_history=result_history))
+        self.writing_stage.execute(result_history=result_history,
+                                   cycle_count=cycle_count))
     self._update_status(result_history=result_history)
     if (not isinstance(result_history[-1], BuildResult) or
         not result_history[-1].success):
@@ -124,7 +125,8 @@ def _execute_one_cycle(self, result_history: list[Result],
     # and if it fails then we will return from this cycle and terminate
     # the pipeline.
     result_history.append(
-        self.execution_stage.execute(result_history=result_history))
+        self.execution_stage.execute(result_history=result_history,
+                                     cycle_count=cycle_count))
     self._update_status(result_history=result_history)
     if (not isinstance(result_history[-1], RunResult) or
         not result_history[-1].log_path):
@@ -137,7 +139,8 @@ def _execute_one_cycle(self, result_history: list[Result],
     # pipeline and retry making a harness. If the analysis stage is successful,
     # then we will terminate the pipeline.
     result_history.append(
-        self.analysis_stage.execute(result_history=result_history))
+        self.analysis_stage.execute(result_history=result_history,
+                                    cycle_count=cycle_count))
     # TODO(maoyi): add the indicator for the success of analysis stage
     if not isinstance(result_history[-1], AnalysisResult):
       self.logger.warning(

report/export.py

@@ -18,6 +18,7 @@
 from abc import abstractmethod
 
 from report.common import Results
+from report.parse_logs import RunLogsParser
 
 
 class BaseExporter:
@@ -42,6 +43,10 @@ def _get_full_url(self, relative_path: str) -> str:
       return relative_path
     return f"{self._base_url}/{relative_path}"
 
+  def get_url_path(self) -> str:
+    """Get the URL path to the CSV file."""
+    return os.path.join(self._output_dir, 'crashes.csv')
+
 
 class CSVExporter(BaseExporter):
   """Export a report to CSV."""
@@ -65,9 +70,19 @@ def generate(self):
 
         project_name = benchmark_id.split("-")[1]
 
+        project_name = benchmark_id.split("-")[1]
+
         for sample in samples:
+          run_logs = self._results.get_run_logs(benchmark_id, sample.id) or ""
+          parser = RunLogsParser(run_logs, benchmark_id, sample.id)
+          crash_reproduction_path = parser.get_crash_reproduction_path()
+
           report_url = self._get_full_url(
               f"sample/{benchmark_id}/{sample.id}.html")
+          reproducer_path = self._get_full_url(
+              f'results/{benchmark_id}/artifacts/{sample.id}.fuzz_target-F0-01/'
+              f'{crash_reproduction_path}') if crash_reproduction_path else ""
+
           writer.writerow({
               "Project":
                   project_name,
@@ -87,12 +102,8 @@ def generate(self):
               "Line Coverage Diff":
                   sample.result.line_coverage_diff,
               "Reproducer Path":
-                  sample.result.reproducer_path
+                  reproducer_path
           })
 
     logging.info("Created CSV file at %s", csv_path)
     return csv_path
-
-  def get_url_path(self):
-    """Get the URL path to the CSV file."""
-    return os.path.join(self._output_dir, 'crashes.csv')