[amartya1523]

Fixes item 1 of tscircuit/tscircuit#5559.

Problem

Autorouted vias and via-stitching vias already use the board's min_via_hole_diameter and min_via_pad_diameter. Hand-placed vias did not, and always used the 0.2/0.3mm defaults. On a board declaring minViaHoleDiameter={0.3} minViaPadDiameter={0.6}, every one of these came out 0.2/0.3mm, a 0.05mm ring that JLCDFM flags as Danger:

• pcbPath / pcbPaths vias
• <pcbtrace> route vias without a size
• <via> with no holeDiameter / outerDiameter
• vias from trace-level routing

A board that mixes autorouted and hand-routed nets therefore ended up with two via sizes.

Changes

• I moved the via-minimum resolution that Board writes onto pcb_board into Board._getResolvedMinViaDiameters(). The order is unchanged: minViaHoleDiameter/minViaPadDiameter, then the board's pcbStyle via size, then JLCPCB's minimums, with the pad derived from the hole when only the hole minimum is given.
• getViaDiameterDefaultsForComponent(component, overrides) is now used by <via>, <pcbtrace>, pcbPath/pcbPaths (both manual-trace render paths) and trace-level routing. The precedence is:
  1. an explicit size (<via> props, or route-point hole_diameter/outer_diameter)
  2. the inherited pcbStyle via size, so a group-level style still wins
  3. the board's resolved via minimums
  4. the 0.2/0.3mm defaults
• Boards without via minimums resolve to JLCPCB's 0.2/0.3mm, the same as the old defaults, so existing designs don't change.

Tests

One test per file, each with a PCB snapshot labelled by pcbnotetext:

• repro5559-hand-placed-vias-board-minimums: pcbPath, <pcbtrace> and <via> on a 0.3/0.6 board. This fails on main with all four vias at 0.2/0.3 and passes with this change.
• via-board-minimum-diameters1: explicit, pcbStyle and route-point sizes still win over the board minimums.
• via-board-minimum-diameters2: with only minViaHoleDiameter, vias match pcb_board.min_via_* (0.3/0.45).
• via-board-minimum-diameters3: a board without minimums keeps 0.2/0.3.

I also ran every test file that touches <via>, pcbPath, <pcbtrace> or via sizes (99 files). Their failures are identical to main on the same machine: pre-existing macOS-vs-CI snapshot render differences, and none are new. tsc --noEmit is clean.

The snapshots were generated on macOS. If CI renders them differently, /update-snapshots should refresh them.

Related @tscircuit/checks PRs for the DFM checks in the same issue: tscircuit/checks#452 (via sizes / annular rings) and tscircuit/checks#453 (trace width).

[vercel]

The latest updates on your projects. Learn more about Vercel for GitHub.

Project	Deployment	Actions	Updated
tscircuit-core-benchmarks	Ready	Preview	Oct 27, 2026 5:25am UTC

[github-actions]

This PR has been automatically marked as stale because it has had no recent activity. It will be closed if no further activity occurs.