ci: bound backend test step summaries (#3410)

Co-authored-by: ayattara <alassane.yattara@savoirfairelinux.com> Co-authored-by: Carlos Florencio <carlosmflorencio@gmail.com>
2026-09-06 06:35:38 -04:00
parent 9dea4b1759
commit 698ef63e67
6 changed files with 373 additions and 7 deletions
@@ -0,0 +1,77 @@
+#!/usr/bin/env python3
+"""Write generated Markdown within GitHub Actions' step-summary size limit."""
+
+import argparse
+import re
+from pathlib import Path
+
+
+FAILURE_IDENTITY = re.compile(r"<li>🔴<code>([^<]*)</code></li>")
+
+
+def parse_args() -> argparse.Namespace:
+    parser = argparse.ArgumentParser()
+    parser.add_argument("--input", type=Path, required=True)
+    parser.add_argument("--output", type=Path, required=True)
+    parser.add_argument("--diagnostics-label", required=True)
+    parser.add_argument("--diagnostics-url", required=True)
+    parser.add_argument("--max-bytes", type=int, default=983_040)
+    return parser.parse_args()
+
+
+def bounded_summary(
+    source: bytes,
+    *,
+    diagnostics_label: str,
+    diagnostics_url: str,
+    max_bytes: int,
+) -> bytes:
+    source_text = source.decode("utf-8")
+    if len(source) <= max_bytes:
+        return source
+
+    notice = (
+        "> [!WARNING]\n"
+        "> This step summary was truncated to stay below GitHub Actions' "
+        "1 MiB limit. Full diagnostics are available in the "
+        f"[{diagnostics_label}]({diagnostics_url}) artifact.\n\n"
+    ).encode("utf-8")
+    if len(notice) > max_bytes:
+        raise ValueError("max bytes is too small for the truncation notice")
+
+    prefix_budget = max_bytes - len(notice)
+    failure_identities = FAILURE_IDENTITY.findall(source_text)
+    identity_section = b""
+    if failure_identities:
+        identity_section = (
+            "## Failed test identities\n\n"
+            + "\n".join(f"- `{identity}`" for identity in failure_identities)
+            + "\n\n"
+        ).encode("utf-8")
+        identity_section = (
+            identity_section[: max(0, prefix_budget)]
+            .decode("utf-8", errors="ignore")
+            .encode("utf-8")
+        )
+
+    prefix_budget -= len(identity_section)
+    # The full source is valid UTF-8, so ignoring errors can only discard a
+    # partial code point at the byte boundary.
+    prefix = source[:prefix_budget].decode("utf-8", errors="ignore").encode("utf-8")
+    return notice + identity_section + prefix
+
+
+def main() -> int:
+    args = parse_args()
+    output = bounded_summary(
+        args.input.read_bytes(),
+        diagnostics_label=args.diagnostics_label,
+        diagnostics_url=args.diagnostics_url,
+        max_bytes=args.max_bytes,
+    )
+    args.output.write_bytes(output)
+    return 0
+
+
+if __name__ == "__main__":
+    raise SystemExit(main())