fix: strip the five surviving +1 emit sites (literal search, imports)

src/mcp/literal_search.rs

@@ -135,7 +135,7 @@ impl CodesearchService {
                             &effective_query,
                             snippet_regex.as_ref(),
                         ) {
-                            let match_line = chunk.start_line + match_offset + 1;
+                            let match_line = chunk.start_line + match_offset;
                             items.push(LiteralSearchResultItem {
                                 path: ctx.prefix_sourced_path(alias, &chunk.path),
                                 start_line: match_line,
@@ -189,7 +189,7 @@ impl CodesearchService {
                                     &effective_query,
                                     snippet_regex.as_ref(),
                                 ) {
-                                    let match_line = chunk.start_line + match_offset + 1;
+                                    let match_line = chunk.start_line + match_offset;
                                     items.push(LiteralSearchResultItem {
                                         path: chunk.path,
                                         start_line: match_line,
@@ -423,7 +423,7 @@ impl CodesearchService {
                         let (match_offset, snippet) = match_info.unwrap_or_else(|| {
                             (0, chunk.content.lines().next().unwrap_or("").to_string())
                         });
-                        let match_line = chunk.start_line + match_offset + 1;
+                        let match_line = chunk.start_line + match_offset;
                         items.push(LiteralSearchResultItem {
                             path: ctx.prefix_sourced_path(&fts_hit.alias, &chunk.path),
                             start_line: match_line,
@@ -506,7 +506,7 @@ impl CodesearchService {
                                     let (match_offset, snippet) = match_info.unwrap_or_else(|| {
                                         (0, chunk.content.lines().next().unwrap_or("").to_string())
                                     });
-                                    let match_line = chunk.start_line + match_offset + 1;
+                                    let match_line = chunk.start_line + match_offset;
                                     Some(LiteralSearchResultItem {
                                         path: chunk.path,
                                         start_line: match_line,

src/mcp/mod.rs

@@ -977,7 +977,7 @@ fn parse_import_lines(content: &str, start_line: usize) -> Vec<ImportItem> {
                 imported,
                 // `start_line` is the chunk's 0-based first line and `offset`
                 // is 0-based within it; the response emits 1-based lines.
-                line: start_line + offset + 1,
+                line: start_line + offset,
                 kind,
             });
         }

src/mcp/search.rs

@@ -1320,8 +1320,8 @@ impl CodesearchService {
             .map(|r| SearchResultItem {
                 chunk_id: Some(r.id),
                 path: r.path,
-                // Chunks store 0-indexed lines; every response emits 1-based
-                // editor lines, so the conversion happens once, here.
+                // Chunks store 1-based editor lines; responses pass them
+                // through unchanged.
                 start_line: r.start_line,
                 end_line: r.end_line,
                 kind: r.kind,

src/mcp/tests.rs

@@ -1220,19 +1220,21 @@ fn test_match_line_for_literal_regex() {
 #[test]
 fn test_parse_import_lines_detects_common_forms() {
     let content = "use std::fs;\nimport os\nfrom pkg import thing\n#include <stdio.h>\nconst x = require('x')\nlet y = 1;";
-    // `10` is the chunk's 0-based first line; emitted lines are 1-based.
+    // `10` is the chunk's 1-based first line (the chunk starts on editor
+    // line 10); match lines are start_line + the 0-based offset within the
+    // chunk.
     let imports = super::parse_import_lines(content, 10);
     assert_eq!(imports.len(), 5);
     assert_eq!(imports[0].kind, "use");
-    assert_eq!(imports[0].line, 11);
+    assert_eq!(imports[0].line, 10);
     assert_eq!(imports[1].kind, "import");
-    assert_eq!(imports[1].line, 12);
+    assert_eq!(imports[1].line, 11);
     assert_eq!(imports[2].kind, "import");
-    assert_eq!(imports[2].line, 13);
+    assert_eq!(imports[2].line, 12);
     assert_eq!(imports[3].kind, "include");
-    assert_eq!(imports[3].line, 14);
+    assert_eq!(imports[3].line, 13);
     assert_eq!(imports[4].kind, "require");
-    assert_eq!(imports[4].line, 15);
+    assert_eq!(imports[4].line, 14);
 }
 
 // === Project/group routing tests ===

src/serve/tests.rs

@@ -5936,14 +5936,14 @@ async fn group_literal_search_relaxed_fallback_covers_partial_term_hits() {
             let mut vs = stores.vector_store.write().await;
             let rich = crate::chunker::Chunk::new(
                 "create package documents quickly".to_string(),
-                0,
+                1,
                 3,
                 crate::chunker::ChunkKind::Comment,
                 "docs/create.md".to_string(),
             );
             let poor = crate::chunker::Chunk::new(
                 "package".to_string(),
-                0,
+                1,
                 1,
                 crate::chunker::ChunkKind::Comment,
                 "docs/other.md".to_string(),
@@ -6016,7 +6016,7 @@ async fn group_literal_search_relaxed_fallback_covers_partial_term_hits() {
     assert_eq!(items[0]["path"], "partial/docs/create.md", "got: {text}");
     assert_eq!(
         items[0]["start_line"], 1,
-        "responses must emit 1-based editor lines (chunk starts at 0-based line 0): {text}"
+        "responses must emit 1-based editor lines (the chunk starts on editor line 1): {text}"
     );
     assert_eq!(
         parsed["relaxed_fallback"],