Investigate OCC mesh construction SIGSEGV and union failures on detailed multilayer PCB copper

Open
#20 1 comment 0 reactions 0 assignees View on GitHub

Nobody has claimed this yet.

Assessment

Difficulty
5/5
Estimated time
Over a week
Newbie friendliness
30/100
Issue type
Bug
Clarity
Mostly clear
Activity status
Active
Tech stack
typescript

Research direction

Start by checking the clean current main branch against the captured engine base commit and local patch; the issue does not name source files or tests. Then reduce the failure to TSX rendered through core, preserving copper, capacitor contacts, ports, and material interfaces. Done means a bounded native regression reproduces and verifies conforming mesh output or an actionable diagnostic, with a resource-bounded board rerun.

Written by the indexing model from the issue text.

Description

Problem

Detailed multilayer PCB meshing is blocking DDR signal-integrity validation. Native Gmsh/OpenCASCADE CAD construction fails before a tetrahedral mesh is produced, so Palace never runs.

Scope: this was observed in a task-local modified simulate-return-current engine based on commit aece592843df702ca4830d8b5556ce5cb3e17e1f. It has not yet been reproduced on clean current main. The failing experiment includes local material-tiling/passive-capacitor changes; this report does not establish a regression in a released package or a root cause in Gmsh itself.

Board context: https://tscircuit.com/0hmX/am3352-ddr-track-a-local-dfm-wip-20261007

Captured failure

The literal-copper common D5 experiment has 3,183 positive material cells and 24 physical ports, including 22 passive-capacitor ports. Geometry is partitioned per z interval before conforming shared material interfaces and ports.

Environment recorded by the experiment:

  • Linux x86_64 native runtime.
  • Gmsh 4.14.1 / OpenCASCADE 7.8.1.
  • Engine base commit: aece592843df702ca4830d8b5556ce5cb3e17e1f.
  • Local engine patch SHA-256: 2d18472ce5e4b6225ec6ad9521100d9041bd901b19cbe9fd4a4c0f3189d95f04.
  • Input model SHA-256: 651a32347c167911414aa3695857b8d6ebb3865d4f279d698c82caf674a3c5a6.
  • Original board SHA-256: 4ed4c784f20a492ac5736e026cdc62ce9ed35da8656c6f70c9132ef3be7e9700.
  • Captured process exit: 139 / SIGSEGV (signal 11).
  • Elapsed time: 327.73 s.
  • Maximum RSS: 9,258,204 KiB; address-space limit: 10,000,000 KiB; timeout: 900 s.

Last progress record and termination:

{"stage":"fragment_shared_material_interfaces_and_ports","seconds":61.999506363645196,"materialVolumes":3183,"ports":24}
timeout: the monitored command dumped core
Command terminated by signal 11

The launcher's separately captured exit code is authoritative. GNU time also printed a trailing Exit status: 0 despite its signal-11 report; this must not be mistaken for a successful mesh.

A separate, earlier CAD representation also failed with:

Union failed - BOPAlgo_AlertIntersectionFailed

Another partitioned attempt timed out after 900 s at fragment_air_materials_ports, with 1,685 copper volumes / 7 dielectric volumes / 19 ports and maximum RSS 9,320,312 KiB. These are different attempts; they should not be combined into a single purported stack trace.

Expected behavior

Native TSX-generated PCB geometry should produce a conforming mesh with correct material boundaries, physical ports and copper-interface ownership, or fail with an actionable geometry/resource diagnostic. A native crash or unbounded CAD Boolean stage prevents reliable simulation.

Reproduction and fix work

The captured board experiment is evidence, not yet a portable minimal reproduction. A separate coding task will:

  1. Check clean current main and distinguish upstream behavior from defects introduced by the local experimental engine.
  2. Reduce the failing case to actual TSX source rendered through core, including the new simulation elements. Generated Circuit JSON may be an output, but do not author/import a hand-written Circuit JSON fixture.
  3. Prefer two stacked PRs: first the reproduction (visual geometry/mesh output where useful), then the fix on top.
  4. Preserve physical copper, real ground returns, capacitor contacts, ports and material interfaces. Do not make the test pass by deleting difficult geometry, weakening checks, or silently changing the physical model.
  5. Add bounded regression coverage and diagnostics; verify a small native fixture before a resource-bounded board-case rerun.

Working hypothesis and limits

The failure location implicates CAD interface/Boolean construction on detailed copper. High memory use is observed, but the evidence does not establish whether the cause is an address-space allocation failure, a geometric degeneracy, or another OCC failure. Obtain a reduced reproduction and suitable native diagnostics before claiming the cause.

No accepted board field solution, DDR eye, or DDR pass/fail conclusion exists from these failed runs.

Dominant language
TypeScript
Stars
0
Forks
0
Avg merge
16m
Merged PRs (30d)
1

Getting set up

This project ships no dev container, Dockerfile or contributing guide, so setting up is up to you: start from its README, and see our first-contribution guide for the general steps.

First steps

  1. Read the whole issue, then the project's contributing guide.
  2. Comment on the issue to say you are picking it up — it saves two people doing the same work.
  3. Fork the repository and make your change on a branch.
  4. Open a pull request that references the issue number.

Similar issues

More TypeScript issues

Get new issues in your inbox

A short digest of beginner-friendly GitHub issues.