Prevent oversized Forge PR descriptions from blocking publication (#9014)

* Bound Forge pull request descriptions

* Fix bounded native test diff formatting

Author: jormundur00

forge/docs/git-scripts.md

@@ -75,6 +75,12 @@ parameters its workflow's metrics actually produce. The subsections below state
 that per-publisher subset; publishers whose workflows share a body shape are
 grouped together.
 
+PR bodies must remain publishable through GitHub. The shared publication helper
+therefore bounds optional generated detail below GitHub's body limit while
+preserving the issue link, summary, metrics, intervention record, and local CI
+evidence. Version-to-version test comparisons include a diff stat and a bounded
+excerpt; reviewers use the PR's **Files changed** tab for the complete diff.
+
 ### New library support and coverage improvement
 
 Publishers: `make_pr_new_library_support.py` (`library-new-request`) and
@@ -96,8 +102,8 @@ single-PR run and `Refs:` for non-final chunked dynamic-access chunks
 Publishers: `make_pr_javac_fix.py` (`fixes-javac-fail`) and
 `make_pr_java_run_fix.py` (`fixes-java-run-fail`). They share one body shape:
 the agent generation metrics above, a stats comparison for the bumped version,
-and a unified diff between the previous version's and the new version's test
-sources so reviewers can see exactly what the fix changed. The two differ only
+and a bounded test-source comparison so reviewers can see what the fix changed
+without preventing PR creation. The two differ only
 in workflow identity (compilation vs. runtime wording), the metrics file, and
 the PR label (§WF-java-fail-fix-workflow).
 

forge/git_scripts/make_pr_improve_coverage.py

@@ -28,6 +28,7 @@
     BASE_BRANCH,
     REPO,
     REVIEWERS,
+    bound_pr_body,
     parse_pr_number,
     publish_branch,
 )
@@ -331,7 +332,7 @@ def create_pull_request(
         "gh", "pr", "create",
         "--repo", REPO,
         "--title", title,
-        "--body", body,
+        "--body", bound_pr_body(body),
         "--base", BASE_BRANCH,
         "--head", f"{origin_owner}:{branch}",
         "--label", "GenAI",

forge/git_scripts/make_pr_java_run_fix.py

@@ -23,7 +23,8 @@
     BASE_BRANCH,
     REPO,
     REVIEWERS,
-    generate_diff_text,
+    bound_pr_body,
+    format_bounded_test_diff_section,
     publish_branch,
     stage_library_version_paths,
 )
@@ -88,7 +89,7 @@ def create_pull_request(
         "--title",
         title,
         "--body",
-        body,
+        bound_pr_body(body),
         "--base",
         BASE_BRANCH,
         "--head",
@@ -158,11 +159,7 @@ def build_pull_request_preview(
 
 {format_forge_revision_section()}
 {format_stats_diff(repo_path, old_coordinates, new_coordinates)}
-**Comparison between existing test version and AI-Generated update**
-
-```diff
-{generate_diff_text(group, artifact, old_version, new_version, repo_path)}
-```
+{format_bounded_test_diff_section(group, artifact, old_version, new_version, repo_path)}
 """
     post_generation_intervention = metrics_entry.get("post_generation_intervention")
     if post_generation_intervention:

forge/git_scripts/make_pr_javac_fix.py

@@ -23,7 +23,8 @@
     BASE_BRANCH,
     REPO,
     REVIEWERS,
-    generate_diff_text,
+    bound_pr_body,
+    format_bounded_test_diff_section,
     publish_branch,
     stage_library_version_paths,
 )
@@ -93,7 +94,7 @@ def create_pull_request(
         "--title",
         title,
         "--body",
-        body,
+        bound_pr_body(body),
         "--base",
         BASE_BRANCH,
         "--head",
@@ -163,11 +164,7 @@ def build_pull_request_preview(
 
 {format_forge_revision_section()}
 {format_stats_diff(repo_path, old_coordinates, new_coordinates)}
-**Comparison between existing test version and AI-Generated update**
-
-```diff
-{generate_diff_text(group, artifact, old_version, new_version, repo_path)}
-```
+{format_bounded_test_diff_section(group, artifact, old_version, new_version, repo_path)}
 """
     post_generation_intervention = metrics_entry.get("post_generation_intervention")
     if post_generation_intervention:

forge/git_scripts/make_pr_new_library_support.py

@@ -28,6 +28,7 @@
     BASE_BRANCH,
     REPO,
     REVIEWERS,
+    bound_pr_body,
     parse_pr_number,
     publish_branch,
     stage_library_version_paths,
@@ -405,7 +406,7 @@ def create_pull_request(
         "gh", "pr", "create",
         "--repo", REPO,
         "--title", title,
-        "--body", body,
+        "--body", bound_pr_body(body),
         "--base", BASE_BRANCH,
         "--head", f"{origin_owner}:{branch}",
         "--label", "GenAI",

forge/git_scripts/make_pr_ni_run_fix.py

@@ -25,7 +25,8 @@
     BASE_BRANCH,
     REPO,
     REVIEWERS,
-    generate_diff_text,
+    bound_pr_body,
+    format_bounded_test_diff_section,
     publish_branch,
 )
 from utility_scripts.metadata_index import resolve_test_version
@@ -170,7 +171,7 @@ def create_pull_request(
         "gh", "pr", "create",
         "--repo", REPO,
         "--title", title,
-        "--body", body,
+        "--body", bound_pr_body(body),
         "--base", BASE_BRANCH,
         "--head", f"{origin_owner}:{branch}",
         "--label", pr_label,
@@ -220,12 +221,7 @@ def build_test_comparison_section(group: str, artifact: str, old_version: str, n
     new_test_dir = os.path.join(repo_path, "tests", "src", group, artifact, new_version)
     if not os.path.isdir(new_test_dir):
         return ""
-    return (
-        "\n\n**Comparison between existing test version and AI-Generated update**\n\n"
-        "```diff\n"
-        f"{generate_diff_text(group, artifact, old_version, new_version, repo_path)}\n"
-        "```"
-    )
+    return format_bounded_test_diff_section(group, artifact, old_version, new_version, repo_path)
 
 
 def build_pull_request_preview(

forge/git_scripts/make_pr_not_for_native_image.py

@@ -22,6 +22,7 @@
     BASE_BRANCH,
     REPO,
     REVIEWERS,
+    bound_pr_body,
     publish_branch,
 )
 from utility_scripts.metadata_index import get_not_for_native_image_marker
@@ -100,7 +101,7 @@ def create_pull_request(
         "gh", "pr", "create",
         "--repo", REPO,
         "--title", title,
-        "--body", body,
+        "--body", bound_pr_body(body),
         "--base", BASE_BRANCH,
         "--head", f"{origin_owner}:{branch}",
         "--label", "GenAI",

forge/git_scripts/pr_publication.py

@@ -51,6 +51,9 @@
     "human-intervention-logs",
     "forge/human-intervention-logs",
 )
+MAX_PR_BODY_CHARS: int = 60_000
+MAX_INLINE_TEST_DIFF_CHARS: int = 12_000
+PR_BODY_REQUIRED_TAIL_CHARS: int = 12_000
 
 
 def _publication_resume_marker(repo_path: str) -> ContinuationMarker | None:
@@ -260,8 +263,14 @@ def _copy_git_files_under(repo_path: str, source_dir: str, destination_dir: str)
         shutil.copy2(source_file, destination_file)
 
 
-def generate_diff_text(group: str, artifact: str, current_version: str, new_version: str, repo_path: str) -> str:
-    """Build a diff text between the current and new test directories for a library version."""
+def _generate_test_diff_outputs(
+        group: str,
+        artifact: str,
+        current_version: str,
+        new_version: str,
+        repo_path: str,
+) -> tuple[str, str]:
+    """Build full and stat test diffs for bounded PR descriptions (§GIT-pr-body)."""
     old_dir = os.path.join(repo_path, "tests", "src", group, artifact, current_version)
     new_dir = os.path.join(repo_path, "tests", "src", group, artifact, new_version)
     old_display_dir = os.path.relpath(old_dir, repo_path).replace(os.sep, "/")
@@ -277,23 +286,83 @@ def generate_diff_text(group: str, artifact: str, current_version: str, new_vers
         _copy_git_files_under(repo_path, old_dir, tmp_old)
         _copy_git_files_under(repo_path, new_dir, tmp_new)
 
-        # Compute diff text using: git diff --no-index --find-renames tmp_old tmp_new
+        diff_args = ["git", "diff", "--no-index", "--find-renames"]
         res = subprocess.run(
-            ["git", "diff", "--no-index", "--find-renames", tmp_old, tmp_new],
+            [*diff_args, tmp_old, tmp_new],
             capture_output=True,
             text=True,
         )
         if res.returncode in (0, 1):
             diff_text = res.stdout
+            stat_res = subprocess.run(
+                [*diff_args, "--stat", tmp_old, tmp_new],
+                capture_output=True,
+                text=True,
+            )
+            if stat_res.returncode not in (0, 1):
+                raise RuntimeError(
+                    f"Subprocess failed with return code: {stat_res.returncode}. "
+                    f"Error output:\n{stat_res.stderr.strip()}"
+                )
 
             diff_text = diff_text.replace(tmp_old.replace(os.sep, "/"), old_diff_display_dir)
             diff_text = diff_text.replace(tmp_new.replace(os.sep, "/"), new_diff_display_dir)
             diff_text = diff_text.replace(tmp_old, old_diff_display_dir)
             diff_text = diff_text.replace(tmp_new, new_diff_display_dir)
-            return diff_text
+            return diff_text, stat_res.stdout
 
         error_message = (
             f"Subprocess failed with return code: {res.returncode}. "
             f"Error output:\n{res.stderr.strip()}"
         )
         raise RuntimeError(error_message)
+
+
+def generate_diff_text(group: str, artifact: str, current_version: str, new_version: str, repo_path: str) -> str:
+    """Build the complete test diff for callers that need it."""
+    diff_text, _diff_stat = _generate_test_diff_outputs(
+        group, artifact, current_version, new_version, repo_path,
+    )
+    return diff_text
+
+
+def format_bounded_test_diff_section(
+        group: str,
+        artifact: str,
+        current_version: str,
+        new_version: str,
+        repo_path: str,
+) -> str:
+    """Format a reviewable test diff excerpt that cannot exhaust a PR body (§GIT-pr-body)."""
+    diff_text, diff_stat = _generate_test_diff_outputs(
+        group, artifact, current_version, new_version, repo_path,
+    )
+    excerpt = diff_text[:MAX_INLINE_TEST_DIFF_CHARS]
+    truncation_note = ""
+    if len(diff_text) > MAX_INLINE_TEST_DIFF_CHARS:
+        truncation_note = (
+            "\n\nThe complete test diff is available in this pull request's "
+            "**Files changed** tab."
+        )
+    return (
+        "**Test-source comparison**\n\n"
+        "```text\n"
+        f"{diff_stat.strip()}\n"
+        "```\n\n"
+        "```diff\n"
+        f"{excerpt}\n"
+        "```"
+        f"{truncation_note}"
+    )
+
+
+def bound_pr_body(body: str) -> str:
+    """Keep all publisher PR bodies below GitHub's limit (§GIT-pr-body)."""
+    if len(body) <= MAX_PR_BODY_CHARS:
+        return body
+    truncation_notice = (
+        "\n\n---\n\nAdditional generated detail was omitted to keep this pull request "
+        "description within GitHub's size limit."
+    )
+    head_chars = MAX_PR_BODY_CHARS - len(truncation_notice) - PR_BODY_REQUIRED_TAIL_CHARS
+    return body[:head_chars].rstrip() + truncation_notice + body[-PR_BODY_REQUIRED_TAIL_CHARS:]

forge/tests/test_make_pr_ni_run_fix.py

@@ -62,6 +62,33 @@ def fake_gh(*args: str, check: bool = True):
 
 
 class NativeImageRunFinalizationTests(unittest.TestCase):
+    def test_build_test_comparison_section_returns_bounded_markdown_section(self) -> None:
+        with tempfile.TemporaryDirectory() as repo_path:
+            new_test_dir = os.path.join(
+                repo_path,
+                "tests",
+                "src",
+                "org.example",
+                "demo",
+                "2.0.0",
+            )
+            os.makedirs(new_test_dir)
+            bounded_section = (
+                "**Test-source comparison**\n\n```text\n1 file changed\n```\n\n"
+                "```diff\n+change\n```"
+            )
+
+            with patch.object(
+                    make_pr_ni_run_fix,
+                    "format_bounded_test_diff_section",
+                    return_value=bounded_section,
+            ):
+                comparison_section = make_pr_ni_run_fix.build_test_comparison_section(
+                    "org.example", "demo", "1.0.0", "2.0.0", repo_path,
+                )
+
+        self.assertEqual(comparison_section, bounded_section)
+
     def test_stage_and_commit_includes_jvm_test_source_directories(self) -> None:
         with tempfile.TemporaryDirectory() as repo_path:
             test_version_dir = os.path.join(

forge/tests/test_pr_publication.py

@@ -7,6 +7,7 @@
 import subprocess
 import tempfile
 import unittest
+from unittest.mock import patch
 
 from git_scripts import pr_publication
 from utility_scripts.continuation_marker import CONTINUATION_MARKER_FILENAME
@@ -26,6 +27,34 @@ def _git(repo_path: str, *args: str) -> str:
 
 
 class PrPublicationTests(unittest.TestCase):
+    def test_format_bounded_test_diff_section_truncates_large_diff(self) -> None:
+        diff_text = "Q" * (pr_publication.MAX_INLINE_TEST_DIFF_CHARS + 1)
+        with patch.object(
+                pr_publication,
+                "_generate_test_diff_outputs",
+                return_value=(diff_text, " 1 file changed, 1 insertion(+)")
+        ):
+            section = pr_publication.format_bounded_test_diff_section(
+                "org.example", "demo", "1.0.0", "2.0.0", "/repo",
+            )
+
+        self.assertIn("1 file changed", section)
+        self.assertIn("Files changed", section)
+        self.assertNotIn(diff_text, section)
+        self.assertEqual(section.count("Q"), pr_publication.MAX_INLINE_TEST_DIFF_CHARS)
+
+    def test_bound_pr_body_keeps_body_below_github_limit(self) -> None:
+        oversized_body = (
+            "x" * (pr_publication.MAX_PR_BODY_CHARS + 100)
+            + "\n\n## Local CI Verification\n\nPassed"
+        )
+
+        bounded_body = pr_publication.bound_pr_body(oversized_body)
+
+        self.assertLessEqual(len(bounded_body), pr_publication.MAX_PR_BODY_CHARS)
+        self.assertIn("GitHub's size limit", bounded_body)
+        self.assertIn("Local CI Verification", bounded_body)
+
     def test_preservation_cleanup_keeps_pending_metrics_as_local_input(self) -> None:
         with tempfile.TemporaryDirectory() as repo_path:
             _git(repo_path, "init")