feat(xref): extract cross-references from descriptions, user notes, and fix system note regex

- Fix MENTIONED_RE/CLOSED_BY_RE to match real GitLab format
  ('mentioned in issue #N' / 'mentioned in merge request !N')
- Add GITLAB_URL_RE + parse_url_refs() for full URL extraction
- Add extract_refs_from_descriptions() -> source_method='description_parse'
- Add extract_refs_from_user_notes() -> source_method='note_parse'
- Wire both into orchestrator after system note extraction
- 36 tests: regex fix, URL parsing, integration, idempotency

src/core/note_parser.rs

@@ -22,20 +22,34 @@ pub struct ExtractResult {
     pub parse_failures: usize,
 }
 
+// GitLab system notes include the entity type word: "mentioned in issue #5"
+// or "mentioned in merge request !730". The word is mandatory in real data,
+// but we also keep the old bare-sigil form as a fallback (no data uses it today,
+// but other GitLab instances might differ).
 static MENTIONED_RE: LazyLock<Regex> = LazyLock::new(|| {
     Regex::new(
-        r"mentioned in (?:(?P<project>[\w][\w.\-]*(?:/[\w][\w.\-]*)+))?(?P<sigil>[#!])(?P<iid>\d+)",
+        r"mentioned in (?:issue |merge request )?(?:(?P<project>[\w][\w.\-]*(?:/[\w][\w.\-]*)+))?(?P<sigil>[#!])(?P<iid>\d+)",
     )
     .expect("mentioned regex is valid")
 });
 
 static CLOSED_BY_RE: LazyLock<Regex> = LazyLock::new(|| {
     Regex::new(
-        r"closed by (?:(?P<project>[\w][\w.\-]*(?:/[\w][\w.\-]*)+))?(?P<sigil>[#!])(?P<iid>\d+)",
+        r"closed by (?:issue |merge request )?(?:(?P<project>[\w][\w.\-]*(?:/[\w][\w.\-]*)+))?(?P<sigil>[#!])(?P<iid>\d+)",
     )
     .expect("closed_by regex is valid")
 });
 
+/// Matches full GitLab URLs like:
+/// `https://gitlab.example.com/group/project/-/issues/123`
+/// `https://gitlab.example.com/group/sub/project/-/merge_requests/456`
+static GITLAB_URL_RE: LazyLock<Regex> = LazyLock::new(|| {
+    Regex::new(
+        r"https?://[^\s/]+/(?P<project>[^\s]+?)/-/(?P<entity_type>issues|merge_requests)/(?P<iid>\d+)",
+    )
+    .expect("gitlab url regex is valid")
+});
+
 pub fn parse_cross_refs(body: &str) -> Vec<ParsedCrossRef> {
     let mut refs = Vec::new();
 
@@ -54,6 +68,47 @@ pub fn parse_cross_refs(body: &str) -> Vec<ParsedCrossRef> {
     refs
 }
 
+/// Extract cross-references from GitLab URLs in free-text bodies (descriptions, user notes).
+pub fn parse_url_refs(body: &str) -> Vec<ParsedCrossRef> {
+    let mut refs = Vec::new();
+    let mut seen = std::collections::HashSet::new();
+
+    for caps in GITLAB_URL_RE.captures_iter(body) {
+        let Some(entity_type_raw) = caps.name("entity_type").map(|m| m.as_str()) else {
+            continue;
+        };
+        let Some(iid_str) = caps.name("iid").map(|m| m.as_str()) else {
+            continue;
+        };
+        let Some(project) = caps.name("project").map(|m| m.as_str()) else {
+            continue;
+        };
+        let Ok(iid) = iid_str.parse::<i64>() else {
+            continue;
+        };
+
+        let target_entity_type = match entity_type_raw {
+            "issues" => "issue",
+            "merge_requests" => "merge_request",
+            _ => continue,
+        };
+
+        let key = (target_entity_type, project.to_owned(), iid);
+        if !seen.insert(key) {
+            continue; // deduplicate within same body
+        }
+
+        refs.push(ParsedCrossRef {
+            reference_type: "mentioned".to_owned(),
+            target_entity_type: target_entity_type.to_owned(),
+            target_iid: iid,
+            target_project_path: Some(project.to_owned()),
+        });
+    }
+
+    refs
+}
+
 fn capture_to_cross_ref(
     caps: &regex::Captures<'_>,
     reference_type: &str,
@@ -233,6 +288,189 @@ fn resolve_cross_project_entity(
     resolve_entity_id(conn, project_id, entity_type, iid)
 }
 
+/// Extract cross-references from issue and MR descriptions (GitLab URLs only).
+pub fn extract_refs_from_descriptions(conn: &Connection, project_id: i64) -> Result<ExtractResult> {
+    let mut result = ExtractResult::default();
+
+    let mut insert_stmt = conn.prepare_cached(
+        "INSERT OR IGNORE INTO entity_references
+         (project_id, source_entity_type, source_entity_id,
+          target_entity_type, target_entity_id,
+          target_project_path, target_entity_iid,
+          reference_type, source_method, created_at)
+         VALUES (?1, ?2, ?3, ?4, ?5, ?6, ?7, ?8, 'description_parse', ?9)",
+    )?;
+
+    let now = now_ms();
+
+    // Issues with descriptions
+    let mut issue_stmt = conn.prepare_cached(
+        "SELECT id, iid, description FROM issues
+         WHERE project_id = ?1 AND description IS NOT NULL AND description != ''",
+    )?;
+    let issues: Vec<(i64, i64, String)> = issue_stmt
+        .query_map([project_id], |row| {
+            Ok((row.get(0)?, row.get(1)?, row.get(2)?))
+        })?
+        .collect::<std::result::Result<Vec<_>, _>>()?;
+
+    for (entity_id, _iid, description) in &issues {
+        insert_url_refs(
+            conn,
+            &mut insert_stmt,
+            &mut result,
+            project_id,
+            "issue",
+            *entity_id,
+            description,
+            now,
+        )?;
+    }
+
+    // Merge requests with descriptions
+    let mut mr_stmt = conn.prepare_cached(
+        "SELECT id, iid, description FROM merge_requests
+         WHERE project_id = ?1 AND description IS NOT NULL AND description != ''",
+    )?;
+    let mrs: Vec<(i64, i64, String)> = mr_stmt
+        .query_map([project_id], |row| {
+            Ok((row.get(0)?, row.get(1)?, row.get(2)?))
+        })?
+        .collect::<std::result::Result<Vec<_>, _>>()?;
+
+    for (entity_id, _iid, description) in &mrs {
+        insert_url_refs(
+            conn,
+            &mut insert_stmt,
+            &mut result,
+            project_id,
+            "merge_request",
+            *entity_id,
+            description,
+            now,
+        )?;
+    }
+
+    if result.inserted > 0 || result.skipped_unresolvable > 0 {
+        debug!(
+            inserted = result.inserted,
+            unresolvable = result.skipped_unresolvable,
+            "Description cross-reference extraction complete"
+        );
+    }
+
+    Ok(result)
+}
+
+/// Extract cross-references from user (non-system) notes (GitLab URLs only).
+pub fn extract_refs_from_user_notes(conn: &Connection, project_id: i64) -> Result<ExtractResult> {
+    let mut result = ExtractResult::default();
+
+    let mut note_stmt = conn.prepare_cached(
+        "SELECT n.id, n.body, d.noteable_type,
+                COALESCE(d.issue_id, d.merge_request_id) AS entity_id
+         FROM notes n
+         JOIN discussions d ON n.discussion_id = d.id
+         WHERE n.is_system = 0
+           AND n.project_id = ?1
+           AND n.body IS NOT NULL",
+    )?;
+
+    let notes: Vec<(i64, String, String, i64)> = note_stmt
+        .query_map([project_id], |row| {
+            Ok((row.get(0)?, row.get(1)?, row.get(2)?, row.get(3)?))
+        })?
+        .collect::<std::result::Result<Vec<_>, _>>()?;
+
+    if notes.is_empty() {
+        return Ok(result);
+    }
+
+    let mut insert_stmt = conn.prepare_cached(
+        "INSERT OR IGNORE INTO entity_references
+         (project_id, source_entity_type, source_entity_id,
+          target_entity_type, target_entity_id,
+          target_project_path, target_entity_iid,
+          reference_type, source_method, created_at)
+         VALUES (?1, ?2, ?3, ?4, ?5, ?6, ?7, ?8, 'note_parse', ?9)",
+    )?;
+
+    let now = now_ms();
+
+    for (_, body, noteable_type, entity_id) in &notes {
+        let source_entity_type = noteable_type_to_entity_type(noteable_type);
+        insert_url_refs(
+            conn,
+            &mut insert_stmt,
+            &mut result,
+            project_id,
+            source_entity_type,
+            *entity_id,
+            body,
+            now,
+        )?;
+    }
+
+    if result.inserted > 0 || result.skipped_unresolvable > 0 {
+        debug!(
+            inserted = result.inserted,
+            unresolvable = result.skipped_unresolvable,
+            "User note cross-reference extraction complete"
+        );
+    }
+
+    Ok(result)
+}
+
+/// Shared helper: parse URL refs from a body and insert into entity_references.
+#[allow(clippy::too_many_arguments)]
+fn insert_url_refs(
+    conn: &Connection,
+    insert_stmt: &mut rusqlite::CachedStatement<'_>,
+    result: &mut ExtractResult,
+    project_id: i64,
+    source_entity_type: &str,
+    source_entity_id: i64,
+    body: &str,
+    now: i64,
+) -> Result<()> {
+    let url_refs = parse_url_refs(body);
+
+    for xref in &url_refs {
+        let target_entity_id = if let Some(ref path) = xref.target_project_path {
+            resolve_cross_project_entity(conn, path, &xref.target_entity_type, xref.target_iid)
+        } else {
+            resolve_entity_id(conn, project_id, &xref.target_entity_type, xref.target_iid)
+        };
+
+        let rows_changed = insert_stmt.execute(rusqlite::params![
+            project_id,
+            source_entity_type,
+            source_entity_id,
+            xref.target_entity_type,
+            target_entity_id,
+            xref.target_project_path,
+            if target_entity_id.is_none() {
+                Some(xref.target_iid)
+            } else {
+                None
+            },
+            xref.reference_type,
+            now,
+        ])?;
+
+        if rows_changed > 0 {
+            if target_entity_id.is_none() {
+                result.skipped_unresolvable += 1;
+            } else {
+                result.inserted += 1;
+            }
+        }
+    }
+
+    Ok(())
+}
+
 #[cfg(test)]
 #[path = "note_parser_tests.rs"]
 mod tests;