# Seve - @Abse subcircuit caching should be much ... Channel: #contributor Source: https://discord.com/channels/1233487248129921135/1533931873997422792 Started: 2026-08-03T20:17:38.011000+00:00 Last activity: 2026-08-14T20:34:52.912000+00:00 ## Seve — 2026-08-03T20:17:37.837000+00:00 <@398467566588264449> subcircuit caching should be much better now ## Abse — 2026-08-05T16:09:36.315000+00:00 <@757706909351411845> https://github.com/tscircuit/footprinter/pull/769 ## Seve — 2026-08-05T16:31:17.800000+00:00 we should definitely have a snapshot for this stuff with the footprint we're targeting ## Seve — 2026-08-05T16:31:33.424000+00:00 i wouldn't have approved it because idk if you targeted a non-QFN component ## Seve — 2026-08-05T16:31:39.586000+00:00 how do i know there's not a better function name ## Seve — 2026-08-05T16:32:22.951000+00:00 the "lr" convention is going to be pervasive if we use it- can you share the chip you were attempting to capture ## Abse — 2026-08-05T16:32:37.339000+00:00 ok one sec ## Abse — 2026-08-05T16:34:22.964000+00:00 <@757706909351411845> C128396 ## Seve — 2026-08-05T16:35:46.459000+00:00 ok let's see how it goes ## Seve — 2026-08-05T16:36:11.745000+00:00 you should always introduce the full name in addition to the alias btw ## Abse — 2026-08-05T18:05:29.353000+00:00 https://github.com/tscircuit/footprinter/pull/770 ## Abse — 2026-08-07T14:08:13.666000+00:00 <@757706909351411845> hay **Repros:** [PR #3018](https://github.com/tscircuit/core/pull/3018) includes a minimal case where a parent trace crosses preserved different-net child copper, and a Game Boy-like case where parent-to-MCU routing exhausts the autorouter iteration limit. **Fix:** [PR #3044](https://github.com/tscircuit/core/pull/3044) routes the MCU through a virtual breakout, preserves breakout-point identity across routing phases, and routes explicitly breakout-owned nets early. The large board now routes without physical headers and passes DRC. ## Abse — 2026-08-07T14:08:53.783000+00:00 can you validate if this is a real bug in the breakout and a fix ## Abse — 2026-08-07T14:32:10.444000+00:00 also check these Here’s a brief Discord-ready update: **JLC5000 footprint improvements** All Footprinter PRs have passing CI and produce 99%+ IoU: - **C2883760 – Staggered SMD pin header:** Discovery ignored the vertical row pitch. Added staggered-row detection and measured `py`. **87.20% → 99.997%** https://github.com/tscircuit/circuit-json-to-footprinter/pull/85 - **C151520 – BFP650 SOT-343:** Uses a wider, inward-shifted emitter pad that standard SOT-343 cannot represent. Added a specific BFP650 variant without changing the standard footprint. **89.13% → 99.966%** https://github.com/tscircuit/footprinter/pull/771 - **C22461789 – LED3510:** Four inline LED pads with wider outer contacts were incorrectly matched as FPC. Added an explicit LED3510 footprint. **89.08% → 99.989%** https://github.com/tscircuit/footprinter/pull/772 - **C78322 – SMD potentiometer:** Existing potentiometer support was through-hole only. Added an SMD potentiometer variant with a wider center pad. **88.34% → 99.998%** https://github.com/tscircuit/footprinter/pull/773 - **C2827688 – SOT-143:** Pin 1 is wider and shifted inward, which symmetric families cannot reproduce. Added an explicit SOT-143 footprint. **88.16% → 100%** https://github.com/tscircuit/footprinter/pull/774 - **C29718 – 0603x4 resistor array:** The end terminals are wider than the inner terminals. Added configurable column pitch and outer-pad height to the existing family. **87.39% → 99.973%** https://github.com/tscircuit/footprinter/pull/775 ## Abse — 2026-08-07T14:32:12.088000+00:00 - **C225714 – G5NB relay:** The sparse four-hole layout has two different hole sizes and cannot be modeled as a pin grid. Added a specific G5NB relay footprint. **87.27% → 99.993%** https://github.com/tscircuit/footprinter/pull/776 - **C2979263 – TOLL-8:** Existing discovery approximated the large drain copper as an offset thermal pad. Added a proper TOLL-8 footprint with leads, bridge, and drain pad. **86.22% → 99.997%** https://github.com/tscircuit/footprinter/pull/777 - **C3130 – Fuse clip:** Uses two rotated plated slots, but existing candidates only supported circular holes. Added a slotted fuse-clip footprint. **86.14% → 100%** https://github.com/tscircuit/footprinter/pull/778 - **C35556 – UFQFPN-20:** Eight corner leads are chamfered polygons that regular QFN/LGA pads cannot reproduce. Added an explicit UFQFPN-20 footprint. **85.75% → 100%** https://github.com/tscircuit/footprinter/pull/779 ## Abse — 2026-08-07T14:54:03.541000+00:00 <@757706909351411845> if it\s not a standard do we always try to make it using parameters on existing fps ? ## Seve — 2026-08-07T14:54:36.502000+00:00 ya igs ## Seve — 2026-08-07T14:55:48.763000+00:00 I dont think footprinter should start supporting manufacturer specific stuff ## Abse — 2026-08-07T14:57:11.220000+00:00 so we should reach a point where we don't want a fp string to make it and just use polygon pads or wtvr ## Seve — 2026-08-07T14:59:20.636000+00:00 Yea the relay one is interesting tho ## Abse — 2026-08-07T17:32:21.146000+00:00 <@757706909351411845> https://github.com/tscircuit/footprinter/pull/781 ## Abse — 2026-08-07T17:49:24.341000+00:00 can u review one more time ty ## Abse — 2026-08-07T17:49:25.908000+00:00 <@757706909351411845> ## Seve — 2026-08-07T17:52:13.125000+00:00 i think we should probably accept the smtpads alias for this, also it's driving me crazy that you keep adding "centerpadwidth", which is objectively incorrect ## Seve — 2026-08-07T17:52:35.370000+00:00 smdpads4_centerpadwidth4 (what is the center pad now?) ## Seve — 2026-08-07T17:52:41.568000+00:00 bad design! ## Seve — 2026-08-07T17:52:58.520000+00:00 actually you know what ## Seve — 2026-08-07T17:53:01.796000+00:00 it's fine ## Seve — 2026-08-07T17:53:14.819000+00:00 we don't normally throw errors in footprints ## Seve — 2026-08-07T17:53:28.346000+00:00 but it's fine ## Seve — 2026-08-07T17:53:34.458000+00:00 i guess it is how a human might describe it ## Abse — 2026-08-08T12:00:20.819000+00:00 <@757706909351411845> https://github.com/tscircuit/core/pull/3081 ## Seve — 2026-08-08T14:27:04.343000+00:00 Nice find ## Abse — 2026-08-09T14:31:12.734000+00:00 <@757706909351411845> https://github.com/tscircuit/core/pull/3087 ## Abse — 2026-08-09T14:42:30.145000+00:00 https://github.com/tscircuit/core/pull/3121 ## Seve — 2026-08-09T16:27:49.367000+00:00 nice ## Abse — 2026-08-10T10:59:47.223000+00:00 <@757706909351411845> https://github.com/tscircuit/core/pull/3123 ## Abse — 2026-08-10T10:59:56.934000+00:00 https://github.com/tscircuit/core/pull/3122 ## Abse — 2026-08-10T11:21:41.108000+00:00 https://github.com/tscircuit/cli/pull/4150 ## Abse — 2026-08-13T18:51:40.115000+00:00 <@757706909351411845> https://github.com/tscircuit/core/pull/3198 ## Abse — 2026-08-14T14:08:21.241000+00:00 <@757706909351411845> https://github.com/tscircuit/core/pull/3198 ## Abse — 2026-08-14T14:09:51.473000+00:00 I switched the repro for the fanout bug , looks like its an srj generation bug ## Abse — 2026-08-14T14:55:19.532000+00:00 <@757706909351411845> https://github.com/tscircuit/core/pull/3212 ## Seve — 2026-08-14T15:03:01.085000+00:00 Your prev repro is still valid ## Seve — 2026-08-14T15:03:14.532000+00:00 Missing pointToConnect iiuc ## Abse — 2026-08-14T15:12:16.105000+00:00 yeah will be the next fix ## Abse — 2026-08-14T15:49:34.652000+00:00 <@757706909351411845> https://github.com/tscircuit/core/pull/3213 ## Abse — 2026-08-14T19:19:09.274000+00:00 <@757706909351411845> https://github.com/tscircuit/core/pull/3219 ## Abse — 2026-08-14T20:34:52.912000+00:00 <@757706909351411845> https://github.com/tscircuit/core/pull/3224