fix(api)!: prevent OSM node ID truncation in match/route annotations

[timpara]

Issue

Fixes #7069. OSM node IDs above 2^34 come back wrong in /match and /route annotations.

Root cause

Two truncation bugs on the node-ID path:

1. Storage: PackedOSMIDs packs OSM IDs into 34 bits, so any ID at or above 2^34 is silently masked. The BOOST_ASSERT guard is compiled out in release.
2. FlatBuffers: Annotation.nodes was [uint] and route_api.hpp used std::vector<uint32_t>, truncating IDs above 2^32 again on the FB path. JSON output was already 64-bit.

Changes

• Widen PackedOSMIDs from 34 to 40 bits (max ~1.1 * 10^12).
• Add checkPackedOSMNodeIdFits() and call it at every push site so overflow throws during osrm-extract instead of masking.
• Change FB Annotation.nodes to [ulong], use std::vector<uint64_t> in route_api.hpp, regenerate fbresult_generated.js.
• Add unit tests for round-trip above 2^34 and the overflow guard.
• Update docs/http.md.

Breaking changes

1. On-disk: .osrm.nbg_nodes layout changes; existing datasets must be re-extracted. Recommend bumping package.json minor before release so the fingerprint rejects stale data.
2. FB wire: Annotation.nodes changed from uint32 to uint64. FB clients must update accessors.

Notes

• clang-format was not available in my environment; please run ./scripts/format.sh if needed.
• No cucumber feature added: the framework auto-assigns small IDs and reworking it for huge IDs is out of scope. Unit tests cover the storage regression; the FB change is exercised by the existing FB cucumber matrix.
• Vector tile output already uses OSMNodeID::value_type (uint64_t) and benefits transparently.

[DennisOSRM]

Thanks for the contribution. Please consider adding a test that covers the changed behavior.

[DennisOSRM]

This needs a formatting fix to run CI checks:

diff --git a/unit_tests/util/packed_vector.cpp b/unit_tests/util/packed_vector.cpp
index 0762c9b..99425bc 100644
--- a/unit_tests/util/packed_vector.cpp
+++ b/unit_tests/util/packed_vector.cpp
@@ -215,13 +215,13 @@ BOOST_AUTO_TEST_CASE(packed_osm_ids_above_2pow34_roundtrip)
     const std::vector<std::uint64_t> inputs = {
         0ull,
         1ull,
-        (1ull << 33),                                          // 2^33
-        (1ull << 34),                                          // 2^34, the old truncation boundary
-        (1ull << 34) + 1,                                      // just above 2^34
-        20'000'000'000ull,                                     // realistic projected OSM node ID
-        (1ull << 35),                                          // 2^35
-        osrm::extractor::MAX_PACKED_OSM_NODE_ID - 1,           // just below the new limit
-        osrm::extractor::MAX_PACKED_OSM_NODE_ID,               // exactly at the new limit
+        (1ull << 33),                                // 2^33
+        (1ull << 34),                                // 2^34, the old truncation boundary
+        (1ull << 34) + 1,                            // just above 2^34
+        20'000'000'000ull,                           // realistic projected OSM node ID
+        (1ull << 35),                                // 2^35
+        osrm::extractor::MAX_PACKED_OSM_NODE_ID - 1, // just below the new limit
+        osrm::extractor::MAX_PACKED_OSM_NODE_ID,     // exactly at the new limit
     };
 
     for (auto raw : inputs)

[Copilot]

Pull request overview

This PR fixes incorrect OSM node IDs in /route and /match annotations when IDs exceed previous packing/serialization limits, by widening internal packed storage and the FlatBuffers representation to preserve 64-bit IDs end-to-end.

Changes:

• Increase PackedOSMIDs storage from 34 to 40 bits and add a hard overflow guard (checkPackedOSMNodeIdFits) at write sites to fail during extraction rather than truncating.
• Change FlatBuffers Annotation.nodes from [uint] to [ulong] and update the RouteAPI FlatBuffers emitter and the generated JS bindings accordingly.
• Add unit tests for round-tripping IDs above 2^34 and for the overflow guard; update HTTP docs to reflect 64-bit node IDs (partially).

Reviewed changes

Copilot reviewed 8 out of 8 changed files in this pull request and generated 2 comments.

Show a summary per file

File	Description
unit_tests/util/packed_vector.cpp	Adds regression + guard tests for PackedOSMIDs behavior with large node IDs.
src/extractor/extractor.cpp	Adds overflow guard call before pushing node IDs into packed storage during extraction.
include/extractor/packed_osm_ids.hpp	Widens packed bit-width and introduces MAX_PACKED_OSM_NODE_ID + checkPackedOSMNodeIdFits().
include/extractor/files.hpp	Adds overflow guard when reading raw NBG nodes and pushing node IDs into packed storage.
include/engine/api/route_api.hpp	Switches FlatBuffers annotation node vector to uint64_t to avoid truncation.
include/engine/api/flatbuffers/route.fbs	Changes FlatBuffers schema for Annotation.nodes to [ulong].
features/support/fbresult_generated.js	Regenerates JS FlatBuffers accessors/builders for 64-bit nodes (BigInt/BigUint64Array).
docs/http.md	Updates docs to mention 64-bit node IDs (but currently only in one place).

---

💡 Add Copilot custom instructions for smarter, more guided reviews. Learn how to get started.