test(eval): add lifecycle regression fixtures

Author: haasonsaas

TODO.md

@@ -141,7 +141,7 @@ This roadmap is derived from deep research into Greptile's public docs, blog, MC
 
 91. [ ] Add eval fixtures for external-context alignment, not just diff-local correctness.
 92. [x] Add eval fixtures for merge-readiness judgments and unresolved-blocker classification.
-93. [ ] Add eval fixtures for addressed-vs-stale finding lifecycle inference.
+93. [x] Add eval fixtures for addressed-vs-stale finding lifecycle inference.
 94. [x] Add eval fixtures for multi-hop graph reasoning across call chains and contract edges.
 95. [ ] Add eval runs that compare single-pass review against agentic loop review.
 96. [ ] Add production replay evals using anonymized accepted/rejected review outcomes.

eval/fixtures/repo_regressions/lifecycle_api_drops_followup_addressed.yml

@@ -0,0 +1,34 @@
+name: repo regression - API drops follow-up addressed outcome
+repo_path: ../../..
+diff: |
+  diff --git a/src/server/api/types.rs b/src/server/api/types.rs
+  index 6666666..deadbeef 100644
+  --- a/src/server/api/types.rs
+  +++ b/src/server/api/types.rs
+  @@ -111,9 +111,7 @@ pub(crate) fn build_api_review_session(
+          comments: session
+              .comments
+              .into_iter()
+              .map(|comment| {
+  -                let addressed_by_follow_up =
+  -                    addressed_by_follow_up_comment_ids.contains(&comment.id);
+  -                ApiComment::from_comment(comment, stale_review, addressed_by_follow_up)
+  +                ApiComment::from_comment(comment, stale_review, false)
+              })
+              .collect(),
+          summary: session.summary,
+expect:
+  must_find:
+    - file: src/server/api/types.rs
+      contains_any:
+        - API review sessions discard addressed-by-follow-up state
+        - comments touched by later commits will no longer surface as addressed
+        - stale review responses can report fixed open findings as merely stale
+      rule_id: bug.lifecycle.api-drops-followup-addressed
+  must_not_find:
+    - contains: style
+  summary:
+    merge_readiness: NeedsAttention
+    min_open_blockers: 1
+  min_total: 1
+  max_total: 8

eval/fixtures/repo_regressions/lifecycle_context_only_addressed.yml

@@ -0,0 +1,34 @@
+name: repo regression - context-only edits marked addressed
+repo_path: ../../..
+diff: |
+  diff --git a/src/core/comment/outcomes.rs b/src/core/comment/outcomes.rs
+  index 4444444..deadbeef 100644
+  --- a/src/core/comment/outcomes.rs
+  +++ b/src/core/comment/outcomes.rs
+  @@ -37,9 +37,7 @@ pub fn infer_addressed_by_follow_up_comments(
+          for hunk in &diff.hunks {
+              for line in &hunk.changes {
+  -                if line.change_type != ChangeType::Context {
+  -                    if let Some(old_line_no) = line.old_line_no {
+  -                        changed_old_lines.insert(old_line_no);
+  -                    }
+  +                if let Some(old_line_no) = line.old_line_no {
+  +                    changed_old_lines.insert(old_line_no);
+                  }
+              }
+          }
+expect:
+  must_find:
+    - file: src/core/comment/outcomes.rs
+      contains_any:
+        - context-only lines are being treated as addressed follow-up edits
+        - inserting lines above a finding can falsely mark it addressed
+        - follow-up inference should only count non-context old lines
+      rule_id: bug.lifecycle.context-only-addressed
+  must_not_find:
+    - contains: style
+  summary:
+    merge_readiness: NeedsAttention
+    min_open_blockers: 1
+  min_total: 1
+  max_total: 8

eval/fixtures/repo_regressions/lifecycle_not_addressed_persistence_inversion.yml

@@ -0,0 +1,32 @@
+name: repo regression - persistent findings dropped from not-addressed inference
+repo_path: ../../..
+diff: |
+  diff --git a/src/core/comment/outcomes.rs b/src/core/comment/outcomes.rs
+  index 5555555..deadbeef 100644
+  --- a/src/core/comment/outcomes.rs
+  +++ b/src/core/comment/outcomes.rs
+  @@ -74,8 +74,8 @@ pub fn infer_follow_up_comment_resolution_outcomes(
+      let not_addressed_comment_ids = previous_comments
+          .iter()
+          .filter(|comment| comment.status == CommentStatus::Open)
+          .filter(|comment| !addressed_comment_ids.contains(&comment.id))
+  -        .filter(|comment| current_comment_ids.contains(comment.id.as_str()))
+  +        .filter(|comment| !current_comment_ids.contains(comment.id.as_str()))
+          .map(|comment| comment.id.clone())
+          .collect();
+  
+expect:
+  must_find:
+    - file: src/core/comment/outcomes.rs
+      contains_any:
+        - repeated findings stop being recorded as not addressed
+        - the persistence check is inverted so surviving comments are dropped
+        - this breaks addressed-vs-still-open lifecycle inference for recurring findings
+      rule_id: bug.lifecycle.not-addressed-persistence-inversion
+  must_not_find:
+    - contains: style
+  summary:
+    merge_readiness: NeedsAttention
+    min_open_blockers: 1
+  min_total: 1
+  max_total: 8

src/commands/eval/fixtures.rs

@@ -318,6 +318,100 @@ expect:
         );
     }
 
+    #[test]
+    fn test_checked_in_lifecycle_context_only_fixture_loads_summary_expectations() {
+        let fixture_path = std::path::Path::new(env!("CARGO_MANIFEST_DIR"))
+            .join("eval/fixtures/repo_regressions/lifecycle_context_only_addressed.yml");
+
+        let fixtures = load_eval_fixtures_from_path(&fixture_path).unwrap();
+
+        assert_eq!(fixtures.len(), 1);
+        assert_eq!(
+            fixtures[0].fixture.name.as_deref(),
+            Some("repo regression - context-only edits marked addressed")
+        );
+        assert_eq!(
+            fixtures[0].fixture.expect.must_find[0].rule_id.as_deref(),
+            Some("bug.lifecycle.context-only-addressed")
+        );
+        assert_eq!(
+            fixtures[0]
+                .fixture
+                .expect
+                .summary
+                .merge_readiness
+                .as_deref(),
+            Some("NeedsAttention")
+        );
+        assert_eq!(
+            fixtures[0].fixture.expect.summary.min_open_blockers,
+            Some(1)
+        );
+    }
+
+    #[test]
+    fn test_checked_in_lifecycle_persistence_fixture_loads_summary_expectations() {
+        let fixture_path = std::path::Path::new(env!("CARGO_MANIFEST_DIR")).join(
+            "eval/fixtures/repo_regressions/lifecycle_not_addressed_persistence_inversion.yml",
+        );
+
+        let fixtures = load_eval_fixtures_from_path(&fixture_path).unwrap();
+
+        assert_eq!(fixtures.len(), 1);
+        assert_eq!(
+            fixtures[0].fixture.name.as_deref(),
+            Some("repo regression - persistent findings dropped from not-addressed inference")
+        );
+        assert_eq!(
+            fixtures[0].fixture.expect.must_find[0].rule_id.as_deref(),
+            Some("bug.lifecycle.not-addressed-persistence-inversion")
+        );
+        assert_eq!(
+            fixtures[0]
+                .fixture
+                .expect
+                .summary
+                .merge_readiness
+                .as_deref(),
+            Some("NeedsAttention")
+        );
+        assert_eq!(
+            fixtures[0].fixture.expect.summary.min_open_blockers,
+            Some(1)
+        );
+    }
+
+    #[test]
+    fn test_checked_in_lifecycle_api_fixture_loads_summary_expectations() {
+        let fixture_path = std::path::Path::new(env!("CARGO_MANIFEST_DIR"))
+            .join("eval/fixtures/repo_regressions/lifecycle_api_drops_followup_addressed.yml");
+
+        let fixtures = load_eval_fixtures_from_path(&fixture_path).unwrap();
+
+        assert_eq!(fixtures.len(), 1);
+        assert_eq!(
+            fixtures[0].fixture.name.as_deref(),
+            Some("repo regression - API drops follow-up addressed outcome")
+        );
+        assert_eq!(
+            fixtures[0].fixture.expect.must_find[0].rule_id.as_deref(),
+            Some("bug.lifecycle.api-drops-followup-addressed")
+        );
+        assert_eq!(
+            fixtures[0]
+                .fixture
+                .expect
+                .summary
+                .merge_readiness
+                .as_deref(),
+            Some("NeedsAttention")
+        );
+        assert_eq!(
+            fixtures[0].fixture.expect.summary.min_open_blockers,
+            Some(1)
+        );
+    }
+
     #[test]
     fn test_collect_eval_fixtures_expands_pack_entries_in_sorted_order() {
         let dir = tempdir().unwrap();