fix index backfill start epoch

CHANGELOG.md

@@ -53,6 +53,8 @@
 
 - [#7366](https://github.com/ChainSafe/forest/issues/7366): `forest-cli sync wait -w` no longer leaves a stale `Status:` line behind on each refresh when the report wraps onto more terminal rows than it has lines.
 
+- [#7480](https://github.com/ChainSafe/forest/pull/7480): `forest-tool api serve --index-backfill-epochs N` now backfills N epochs below the chain head. Previously the chain head consumed one of the requested epochs, so only N-1 were indexed.
+
 ## Forest v0.35.0 "Shravan"
 
 Non-mandatory release for all node operators. It includes some fixes and improvements, notably around state-related RPC. Note that this release contains breaking changes, so please read the changelog carefully before upgrading.

src/tool/offline_server/server.rs

@@ -192,7 +192,8 @@ pub async fn start_offline_server(
         backfill_db(
             state_manager,
             &head_ts,
-            RangeSpec::NumTipsets(index_backfill_epochs),
+            // The walk includes the head, so add one to keep it from counting.
+            RangeSpec::NumTipsets(index_backfill_epochs.saturating_add(1)),
         )
         .await?;
     }