fix(‎LibCarla): port lane marking, waypoint loop, lane corners, and pugixml null fixes by youtalk · Pull Request #9652 · carla-simulator/carla

youtalk · 2026-04-07T21:03:46Z

Description

Port of 4 ue4-dev bug fixes:

Lane marking generation
(86fcfe5e8) — Extract ComputeEdgesForLanemark helper to handle zero-width lanes without divide-by-zero. Add edges.first == edges.second guard in tree generation.
Waypoint next/previous infinite loop
(e41ffa03c) — Prevent infinite recursion in GetNext/GetPrevious when two roads with opposite directions are connected as each other's successor.
Lane corners crash (4d4dd1d11) — Clamp s parameter in GetCornerPositions to [0, road_length] to prevent crash when requested s exceeds lane distance.
pugixml null pointer dereference
(4ca87e818 + 6cb77140c) — Add !_root null checks in 6 xml_node insertion methods to prevent crashes on empty nodes. Includes Google Test cases.

Where has this been tested?

Platform(s): Ubuntu 24.04
Python version(s): N/A (C++ only)
Unreal Engine version(s): 5.5

Possible Drawbacks

None. All changes are defensive guards and bug fixes.

This change is

…l null fixes Signed-off-by: Yutaka Kondo <yutaka.kondo@youtalk.jp>

update-docs · 2026-04-07T21:03:52Z

Thanks for opening this pull request! The maintainers of this repository would appreciate it if you would update our CHANGELOG.md based on your changes.

Copilot

Pull request overview

Note

Copilot was unable to run its full agentic suite in this review.

Ports several UE4-dev bug fixes into LibCarla, focusing on preventing crashes/infinite loops and improving robustness in lane marking generation and XML node insertion.

Changes:

Add _root null-guards to multiple pugixml insertion APIs and introduce GTest coverage.
Refactor lane marking generation to use a helper that avoids divide-by-zero on zero-width lanes.
Prevent waypoint traversal infinite recursion and harden lane corner computations (clamping s).

Reviewed changes

Copilot reviewed 6 out of 6 changed files in this pull request and generated 7 comments.

Show a summary per file

File	Description
LibCarla/source/third-party/pugixml/pugixml.cpp	Adds `_root` null checks to avoid crashes during insertion operations.
LibCarla/source/test/common/test_pugixml.cpp	Introduces GTests intended to validate the pugixml null-guard behavior.
LibCarla/source/carla/road/MeshFactory.h	Declares a helper for lane marking edge computation.
LibCarla/source/carla/road/MeshFactory.cpp	Uses the helper to generate lane marking geometry and implements robust edge computation.
LibCarla/source/carla/road/Map.cpp	Adds loop-breaking logic in waypoint traversal and a guard for degenerate edges in tree placement.
LibCarla/source/carla/road/Lane.cpp	Clamps `s` in `GetCornerPositions` to avoid out-of-range crashes.

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

Copilot · 2026-04-07T21:29:33Z

LibCarla/source/carla/road/MeshFactory.cpp

+            std::pair<geom::Vector3D, geom::Vector3D> edges =
+              ComputeEdgesForLanemark(lane_section, lane, s_current, lane_mark_info.width);


This refactor changes behavior for callers that previously used GetCornerPositions(s_current, 0) (e.g., solid lane marking cases) by always using road_param.extra_lane_width inside ComputeEdgesForLanemark. That can shift lane marking geometry relative to the lane border. Consider adding an extra_width (or equivalent) parameter to ComputeEdgesForLanemark and pass the original values (0 vs road_param.extra_lane_width) from each call site.

Good catch. Confirmed: the old Solid case and the final-segment call passed extra_width = 0, while the Broken cases used road_param.extra_lane_width. The refactor unifies them on road_param.extra_lane_width, which silently shifts solid lane marking positions on maps with non-zero extra lane width. To preserve the original behavior exactly, please add an extra_width parameter to ComputeEdgesForLanemark and forward 0 from the call sites that previously used it.

Probably the best action right now is to create a new issue ticket and track this possible issue separately.

I made the issue #9677.

Copilot · 2026-04-07T21:29:34Z

LibCarla/source/carla/road/MeshFactory.cpp

+    std::pair<geom::Vector3D, geom::Vector3D> edges =
+      lane.GetCornerPositions(s_current, road_param.extra_lane_width);


This refactor changes behavior for callers that previously used GetCornerPositions(s_current, 0) (e.g., solid lane marking cases) by always using road_param.extra_lane_width inside ComputeEdgesForLanemark. That can shift lane marking geometry relative to the lane border. Consider adding an extra_width (or equivalent) parameter to ComputeEdgesForLanemark and pass the original values (0 vs road_param.extra_lane_width) from each call site.

Same case as the previous comment.

LibCarla/source/carla/road/MeshFactory.cpp

LibCarla/source/carla/road/Map.cpp

LibCarla/source/test/common/test_pugixml.cpp

LibCarla/source/carla/road/Map.cpp

JesusAnaya

This change has one potential bug (LibCarla/source/carla/road/Map.cpp:1294-1306). Please review my comments.

LibCarla/source/carla/road/Map.cpp

JesusAnaya · 2026-04-08T08:25:17Z

LibCarla/source/carla/road/MeshFactory.cpp

+            std::pair<geom::Vector3D, geom::Vector3D> edges =
+              ComputeEdgesForLanemark(lane_section, lane, s_current, lane_mark_info.width);


Good catch. Confirmed: the old Solid case and the final-segment call passed extra_width = 0, while the Broken cases used road_param.extra_lane_width. The refactor unifies them on road_param.extra_lane_width, which silently shifts solid lane marking positions on maps with non-zero extra lane width. To preserve the original behavior exactly, please add an extra_width parameter to ComputeEdgesForLanemark and forward 0 from the call sites that previously used it.

Probably the best action right now is to create a new issue ticket and track this possible issue separately.

LibCarla/source/carla/road/MeshFactory.cpp

JesusAnaya · 2026-04-08T08:29:48Z

LibCarla/source/carla/road/MeshFactory.cpp

+    std::pair<geom::Vector3D, geom::Vector3D> edges =
+      lane.GetCornerPositions(s_current, road_param.extra_lane_width);


Same case as the previous comment.

LibCarla/source/carla/road/Map.cpp

LibCarla/source/test/common/test_pugixml.cpp

Co-authored-by: Copilot <175728472+Copilot@users.noreply.github.com>

Invert the degenerate-edges guard so that only tree placement is skipped when edges.first == edges.second, instead of using continue which skipped the s_current increment and caused an infinite loop. Signed-off-by: Yutaka Kondo <yutaka.kondo@youtalk.jp>

fix(road): port lane marking, waypoint loop, lane corners, and pugixm…

4ed5370

…l null fixes Signed-off-by: Yutaka Kondo <yutaka.kondo@youtalk.jp>

youtalk closed this Apr 7, 2026

youtalk reopened this Apr 7, 2026

youtalk changed the title ~~fix(road): port lane marking, waypoint loop, lane corners, and pugixml null fixes~~ fix(‎LibCarla): port lane marking, waypoint loop, lane corners, and pugixml null fixes Apr 7, 2026

youtalk marked this pull request as ready for review April 7, 2026 21:26

youtalk requested a review from a team as a code owner April 7, 2026 21:26

Copilot AI review requested due to automatic review settings April 7, 2026 21:26

Copilot AI reviewed Apr 7, 2026

View reviewed changes

youtalk mentioned this pull request Apr 7, 2026

Port ue4-dev features and fixes to ue5-dev (tracking issue) #9635

Open

9 tasks

JesusAnaya reviewed Apr 8, 2026

View reviewed changes

youtalk and others added 2 commits April 8, 2026 16:22

Update LibCarla/source/carla/road/Map.cpp

8f31a84

Co-authored-by: Copilot <175728472+Copilot@users.noreply.github.com>

JesusAnaya approved these changes Apr 9, 2026

View reviewed changes

youtalk mentioned this pull request Apr 14, 2026

ComputeEdgesForLanemark always applies extra_lane_width, shifting solid lane markings #9677

Open

		std::pair<geom::Vector3D, geom::Vector3D> edges =
		ComputeEdgesForLanemark(lane_section, lane, s_current, lane_mark_info.width);

		std::pair<geom::Vector3D, geom::Vector3D> edges =
		lane.GetCornerPositions(s_current, road_param.extra_lane_width);

Uh oh!

Conversation

youtalk commented Apr 7, 2026 • edited by Blyron Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Description

Where has this been tested?

Possible Drawbacks

Uh oh!

update-docs bot commented Apr 7, 2026

Uh oh!

Copilot AI left a comment

Choose a reason for hiding this comment

Pull request overview

Reviewed changes

Uh oh!

Copilot AI Apr 7, 2026

Choose a reason for hiding this comment

Uh oh!

JesusAnaya Apr 8, 2026

Choose a reason for hiding this comment

Uh oh!

youtalk Apr 14, 2026

Choose a reason for hiding this comment

Uh oh!

Copilot AI Apr 7, 2026

Choose a reason for hiding this comment

Uh oh!

JesusAnaya Apr 8, 2026

Choose a reason for hiding this comment

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

JesusAnaya left a comment

Choose a reason for hiding this comment

Uh oh!

Uh oh!

JesusAnaya Apr 8, 2026

Choose a reason for hiding this comment

Uh oh!

Uh oh!

JesusAnaya Apr 8, 2026

Choose a reason for hiding this comment

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

3 participants

youtalk commented Apr 7, 2026 •

edited by Blyron

Loading