| 1 | #!/usr/bin/env python3 |
| 2 | from __future__ import annotations |
| 3 | |
| 4 | import argparse |
| 5 | import json |
| 6 | import shutil |
| 7 | import subprocess # nosec B404 |
| 8 | from dataclasses import asdict, dataclass |
| 9 | from pathlib import Path |
| 10 | |
| 11 | |
| 12 | @dataclass |
| 13 | class CopilotFailure: |
| 14 | failure_class: str |
| 15 | retryable: bool |
| 16 | actionable: bool |
| 17 | diagnostic: str |
| 18 | exit_code: int | None = None |
| 19 | |
| 20 | |
| 21 | TOKEN_PATTERNS = ( |
| 22 | "bad credentials", |
| 23 | "invalid token", |
| 24 | "expired token", |
| 25 | "token expired", |
| 26 | "authentication failed", |
| 27 | "unauthorized", |
| 28 | "copilot_github_token", |
| 29 | ) |
| 30 | INACCESSIBLE_PATTERNS = ( |
| 31 | "401", |
| 32 | "403", |
| 33 | "copilot is not available", |
| 34 | "copilot unavailable", |
| 35 | "permission", |
| 36 | "not subscribed", |
| 37 | "subscription", |
| 38 | "access to copilot", |
| 39 | "forbidden", |
| 40 | ) |
| 41 | CONTEXT_PATTERNS = ( |
| 42 | "context length", |
| 43 | "context too large", |
| 44 | "maximum context", |
| 45 | "token limit", |
| 46 | "too many tokens", |
| 47 | "prompt is too long", |
| 48 | ) |
| 49 | TIMEOUT_PATTERNS = ("timed out", "timeout", "deadline exceeded") |
| 50 | TRANSIENT_PATTERNS = ( |
| 51 | "rate limit", |
| 52 | "429", |
| 53 | "500", |
| 54 | "502", |
| 55 | "503", |
| 56 | "504", |
| 57 | "econnreset", |
| 58 | "network", |
| 59 | "temporarily unavailable", |
| 60 | "temporary failure", |
| 61 | ) |
| 62 | |
| 63 | |
| 64 | def classify_log(log_text: str, exit_code: int | None = None) -> CopilotFailure: |
| 65 | normalized = log_text.lower() |
| 66 | if "copilot is not available" in normalized or "command not found" in normalized: |
| 67 | return CopilotFailure( |
| 68 | "copilot_inaccessible", |
| 69 | retryable=False, |
| 70 | actionable=True, |
| 71 | diagnostic="Copilot CLI is unavailable in the runner; install Copilot CLI or verify runner access.", |
| 72 | exit_code=exit_code, |
| 73 | ) |
| 74 | if any(pattern in normalized for pattern in TOKEN_PATTERNS): |
| 75 | return CopilotFailure( |
| 76 | "copilot_token_failure", |
| 77 | retryable=False, |
| 78 | actionable=True, |
| 79 | diagnostic="Copilot authentication/token failure; renew COPILOT_GH_TOKEN.", |
| 80 | exit_code=exit_code, |
| 81 | ) |
| 82 | if any(pattern in normalized for pattern in CONTEXT_PATTERNS): |
| 83 | return CopilotFailure( |
| 84 | "context_too_large", |
| 85 | retryable=False, |
| 86 | actionable=True, |
| 87 | diagnostic="Copilot prompt/context exceeded supported size; reduce analysis context before rerun.", |
| 88 | exit_code=exit_code, |
| 89 | ) |
| 90 | if any(pattern in normalized for pattern in TIMEOUT_PATTERNS): |
| 91 | return CopilotFailure( |
| 92 | "timeout", |
| 93 | retryable=True, |
| 94 | actionable=False, |
| 95 | diagnostic="Copilot analysis timed out; retry is allowed.", |
| 96 | exit_code=exit_code, |
| 97 | ) |
| 98 | if any(pattern in normalized for pattern in TRANSIENT_PATTERNS): |
| 99 | return CopilotFailure( |
| 100 | "transient_error", |
| 101 | retryable=True, |
| 102 | actionable=False, |
| 103 | diagnostic="Copilot analysis hit a transient service/network failure; retry is allowed.", |
| 104 | exit_code=exit_code, |
| 105 | ) |
| 106 | if any(pattern in normalized for pattern in INACCESSIBLE_PATTERNS): |
| 107 | return CopilotFailure( |
| 108 | "copilot_token_failure", |
| 109 | retryable=False, |
| 110 | actionable=True, |
| 111 | diagnostic="Copilot authentication/access failure; renew COPILOT_GH_TOKEN or verify Copilot permissions.", |
| 112 | exit_code=exit_code, |
| 113 | ) |
| 114 | return CopilotFailure( |
| 115 | "other", |
| 116 | retryable=True, |
| 117 | actionable=False, |
| 118 | diagnostic="Copilot analysis failed with an unclassified error; retry is allowed before failing the run.", |
| 119 | exit_code=exit_code, |
| 120 | ) |
| 121 | |
| 122 | |
| 123 | def issue_title() -> str: |
| 124 | return "Renew GitHub Copilot token for weekly analysis workflow" |
| 125 | |
| 126 | |
| 127 | def issue_body(report: CopilotFailure, *, week: str, run_id: str) -> str: |
| 128 | return "\n".join( |
| 129 | [ |
| 130 | "The weekly analysis workflow cannot run because GitHub Copilot authentication or access failed.", |
| 131 | "", |
| 132 | f"- Week: `{week}`", |
| 133 | f"- Run ID: `{run_id}`", |
| 134 | f"- Failure class: `{report.failure_class}`", |
| 135 | f"- Diagnostic: {report.diagnostic}", |
| 136 | "", |
| 137 | "Please renew or replace the `COPILOT_GH_TOKEN` secret, verify Copilot access, then rerun the workflow.", |
| 138 | ] |
| 139 | ) |
| 140 | |
| 141 | |
| 142 | def run_gh(args: list[str]) -> subprocess.CompletedProcess[str]: |
| 143 | gh_path = shutil.which("gh") |
| 144 | if gh_path is None: |
| 145 | raise RuntimeError("GitHub CLI executable not found on PATH") |
| 146 | return subprocess.run([gh_path, *args], check=False, capture_output=True, text=True) # nosec B603 |
| 147 | |
| 148 | |
| 149 | def issue_url(repo: str, number: str) -> str: |
| 150 | return f"https://github.com/{repo}/issues/{number}" |
| 151 | |
| 152 | |
| 153 | def create_or_update_token_issue( |
| 154 | report: CopilotFailure, *, repo: str, assignee: str, week: str, run_id: str |
| 155 | ) -> str: |
| 156 | title = issue_title() |
| 157 | body = issue_body(report, week=week, run_id=run_id) |
| 158 | search = run_gh( |
| 159 | [ |
| 160 | "issue", |
| 161 | "list", |
| 162 | "--repo", |
| 163 | repo, |
| 164 | "--state", |
| 165 | "open", |
| 166 | "--search", |
| 167 | title, |
| 168 | "--json", |
| 169 | "number,title", |
| 170 | "--limit", |
| 171 | "10", |
| 172 | ] |
| 173 | ) |
| 174 | if search.returncode == 0: |
| 175 | try: |
| 176 | issues = json.loads(search.stdout) |
| 177 | except json.JSONDecodeError: |
| 178 | issues = [] |
| 179 | for issue in issues: |
| 180 | if issue.get("title") == title and issue.get("number"): |
| 181 | number = str(issue["number"]) |
| 182 | comment = run_gh(["issue", "comment", number, "--repo", repo, "--body", body]) |
| 183 | if comment.returncode != 0: |
| 184 | raise RuntimeError( |
| 185 | comment.stderr.strip() or "failed to update Copilot token issue" |
| 186 | ) |
| 187 | return issue_url(repo, number) |
| 188 | |
| 189 | created = run_gh( |
| 190 | [ |
| 191 | "issue", |
| 192 | "create", |
| 193 | "--repo", |
| 194 | repo, |
| 195 | "--title", |
| 196 | title, |
| 197 | "--body", |
| 198 | body, |
| 199 | "--assignee", |
| 200 | assignee, |
| 201 | "--label", |
| 202 | "type:bug", |
| 203 | ] |
| 204 | ) |
| 205 | if created.returncode != 0: |
| 206 | raise RuntimeError(created.stderr.strip() or "failed to create Copilot token issue") |
| 207 | created_output = created.stdout.strip() |
| 208 | if created_output.startswith("https://"): |
| 209 | return created_output |
| 210 | return created_output or issue_url(repo, "unknown") |
| 211 | |
| 212 | |
| 213 | def parse_args(argv: list[str] | None = None) -> argparse.Namespace: |
| 214 | parser = argparse.ArgumentParser(description="Classify Copilot CLI analysis failures.") |
| 215 | parser.add_argument("--log", required=True, type=Path) |
| 216 | parser.add_argument("--exit-code", type=int) |
| 217 | parser.add_argument("--report-json", type=Path) |
| 218 | parser.add_argument("--create-token-issue", action="store_true") |
| 219 | parser.add_argument("--repo", default="") |
| 220 | parser.add_argument("--assignee", default="jmservera") |
| 221 | parser.add_argument("--week", default="") |
| 222 | parser.add_argument("--run-id", default="") |
| 223 | return parser.parse_args(argv) |
| 224 | |
| 225 | |
| 226 | def main(argv: list[str] | None = None) -> int: |
| 227 | args = parse_args(argv) |
| 228 | log_text = args.log.read_text(encoding="utf-8", errors="replace") if args.log.exists() else "" |
| 229 | report = classify_log(log_text, args.exit_code) |
| 230 | payload = asdict(report) |
| 231 | payload["log_path"] = args.log.as_posix() |
| 232 | |
| 233 | if args.create_token_issue and report.failure_class in { |
| 234 | "copilot_token_failure", |
| 235 | "copilot_inaccessible", |
| 236 | }: |
| 237 | payload["issue"] = create_or_update_token_issue( |
| 238 | report, |
| 239 | repo=args.repo, |
| 240 | assignee=args.assignee, |
| 241 | week=args.week, |
| 242 | run_id=args.run_id, |
| 243 | ) |
| 244 | |
| 245 | if args.report_json: |
| 246 | args.report_json.parent.mkdir(parents=True, exist_ok=True) |
| 247 | args.report_json.write_text( |
| 248 | json.dumps(payload, indent=2, sort_keys=True) + "\n", encoding="utf-8" |
| 249 | ) |
| 250 | |
| 251 | print(report.failure_class) |
| 252 | return 0 |
| 253 | |
| 254 | |
| 255 | if __name__ == "__main__": |
| 256 | raise SystemExit(main()) |