# Seve - The traces arent correctly routing in th...

Channel: #contributor
Source: https://discord.com/channels/1233487248129921135/1515328809979740281
Started: 2026-06-13T12:15:42.275000+00:00
Last activity: 2026-06-24T18:35:05.682000+00:00

## Seve — 2026-06-13T12:15:41.747000+00:00

The traces arent correctly routing in this doc with assignable vias, can someone take a look? The traces are missing segments (cc <@809856037376491570> ） https://docs.tscircuit.com/guides/tscircuit-essentials/using-viagrid-prefabricated-vias-with-autorouter

## Abse — 2026-06-13T12:18:52.630000+00:00

[attachment]
Attachment: image.png — https://community.tscircuit.com/media/1515329610345218171

## Abse — 2026-06-13T12:18:56.481000+00:00

which has a messing sgement ?

## Abse — 2026-06-13T12:19:20.070000+00:00

the one in the bottom looks like its doing 180 deg turn

## Rishabh — 2026-06-13T12:19:40.366000+00:00

Yeah, something looks weird. I am taking a look

## Abse — 2026-06-13T12:20:07.765000+00:00

let me know if u want me to take a look

## Rishabh — 2026-06-13T12:38:41.792000+00:00

<@757706909351411845> It's because of effortLevel, I removed the `10x` effortLevel and it solves correctly

## Rishabh — 2026-06-13T12:39:05.620000+00:00

Something with the effortLevel is causing this bug

## Rishabh — 2026-06-13T12:41:23.260000+00:00

I am removing that prop from the doc now, and taking a look into the bug

## Seve — 2026-06-14T03:43:51.694000+00:00

Hmm strange

## Abse — 2026-06-15T13:01:49.871000+00:00

<@809856037376491570> https://github.com/tscircuit/footprinter/pull/659

## Abse — 2026-06-15T13:04:43.903000+00:00

<@757706909351411845>

## Abse — 2026-06-22T09:12:25.546000+00:00

<@809856037376491570> https://github.com/tscircuit/kicad-mod-cache/pull/24

## Abse — 2026-06-22T09:12:55.463000+00:00

I used eval in the dev dep to render the foot print generated from this pr is this a good repro for the issue ?

## Abse — 2026-06-22T09:13:57.593000+00:00

or can u give me a better way if this is not the right way

## Rishabh — 2026-06-22T09:15:34.926000+00:00

Yeah, I don't like this. This test was failing in eval. So, just use that footprint string and snapshot the test output
Attachment: image.png — https://community.tscircuit.com/media/1518544973430259722

## Rishabh — 2026-06-22T09:16:29.012000+00:00

This was gettting flipped, just make sure that footprint's silkscreen text is not getting flipped
Attachment: image.png — https://community.tscircuit.com/media/1518545200245510224

## Abse — 2026-06-22T09:16:32.328000+00:00

so the issue is not the 
kicad-mod-cache ?

## Rishabh — 2026-06-22T09:16:37.254000+00:00

No

## Abse — 2026-06-22T09:16:43.127000+00:00

oh ok mb

## Abse — 2026-06-22T09:16:49.993000+00:00

I will check it rn

## Abse — 2026-06-22T11:08:21.851000+00:00

<@809856037376491570> I managed to repro in eval and this should fix the issue in kicad to circuit json https://github.com/tscircuit/kicad-to-circuit-json/pull/150

## Abse — 2026-06-22T11:08:40.266000+00:00

but not sure if we can push the kicad mod code to our repo ?

## Abse — 2026-06-22T11:08:58.989000+00:00

maybe copyrighted, I can remove it after u take a look

## Abse — 2026-06-22T12:12:47.194000+00:00

<@809856037376491570> wehn u got time plz take a look

## Rishabh — 2026-06-22T13:03:04.487000+00:00

Taking a look

## Rishabh — 2026-06-22T13:05:06.461000+00:00

After you merge the kicad-mod-cache migration PR, also run the test Ci in eval after that

## Abse — 2026-06-22T13:23:13.806000+00:00

OK

## Abse — 2026-06-22T13:23:41.989000+00:00

can we merge the kicad mode code or I should remove it ?

## Rishabh — 2026-06-22T13:25:37.025000+00:00

Remove it I guess

## Abse — 2026-06-22T14:29:04.247000+00:00

k

## Abse — 2026-06-22T15:34:44.150000+00:00

<@809856037376491570> can I get another review on this so we can merge https://github.com/tscircuit/kicad-mod-cache/pull/24

## Abse — 2026-06-22T17:30:27.232000+00:00

<@809856037376491570> https://github.com/tscircuit/kicad-mod-cache/pull/24

## Abse — 2026-06-22T17:38:56.971000+00:00

<@809856037376491570> can I get a reivew ?

## Abse — 2026-06-22T20:07:21.158000+00:00

<@809856037376491570> so the text is bigger now for kicad Ref 

In kicad-to-circuit-json, the footprint text path does:
const fontSize = kicadFontSize * KICAD_TEXT_HEIGHT_TO_CIRCUIT_JSON_FONT_SIZE

`The old kicad-component-converter used the KiCad font size directly:
font_size: fp_text.effects?.font?.size[0] ?? 1`

So a KiCad silkscreen ref text with size 1 used to become font_size: 1, but now becomes font_size: 1.5.
The reason in the new converter is documented in its comment: it is trying to be inverse-compatible with circuit-json-to-kicad, which apparently divides by 1.5 when going the other direction.
Attachment: image.png — https://community.tscircuit.com/media/1518708996817682492

## Abse — 2026-06-22T20:08:33.224000+00:00

should we change kicd to circuit json to use the kicad font size or keep the current 1.5 ?

## Rishabh — 2026-06-22T20:21:04.561000+00:00

I think we should keep the size to be inverse compatible. Not so sure.

<@757706909351411845> ?

## Seve — 2026-06-23T05:22:50.667000+00:00

Hmm, i dont really know whats reasonable here but if you convert kicad to circuit json it should be the right size in the snapshot (after unit conversions)

## Seve — 2026-06-23T05:23:14.626000+00:00

Anything else is strange no?

## Abse — 2026-06-23T12:28:57.391000+00:00

<@809856037376491570> I think I should just make the font as it was before ( that looks better imo)

## Rishabh — 2026-06-23T12:30:10.731000+00:00

but also figure out why the conversion scale is not 1.5 times, this might become an annoying bug

## Abse — 2026-06-23T12:59:30.280000+00:00

why do we sacle the text I'm not sure ?

## Abse — 2026-06-23T12:59:33.681000+00:00

https://github.com/tscircuit/kicad-to-circuit-json/pull/154

## Abse — 2026-06-23T13:03:21.616000+00:00

also can u review <@809856037376491570>

## Rishabh — 2026-06-23T13:09:43.914000+00:00

Scaling text makes sense for the schematic's cause what we see in our schematic viewer and what we see in the schematic's in the A3 size or something then it does needs scaling

## Abse — 2026-06-23T13:37:31.993000+00:00

I see so will this create bugs later for sch ?

## Rishabh — 2026-06-23T13:52:26.800000+00:00

I am not sure, run tests

## Abse — 2026-06-23T15:12:22.169000+00:00

we have another issue <@809856037376491570> 
Some KiCad cache requests return 500 Internal Server Error during footprint conversion.
Example failing URLs:
https://kicad-mod-cache.tscircuit.com/Package_TO_SOT_SMD.pretty/SOT-23-3.circuit.json
https://kicad-mod-cache.tscircuit.com/Package_SO.pretty/SOIC-8_3.9x4.9mm_P1.27mm.circuit.json


should we Update kicad-to-circuit-json so fp_poly parsing does not require uuid or tstamp. If both are missing, generate a deterministic fallback id from the polygon data, Then update the cache service to use the fixed converter ?

## Rishabh — 2026-06-23T15:17:13.021000+00:00

Yeah, repro this inside `kicad-to-circuit-json` and then create a fix for it

## Rishabh — 2026-06-23T15:17:21.639000+00:00

After that update the kicad-mod-cache

## Abse — 2026-06-24T12:52:44.125000+00:00

<@809856037376491570>

## Abse — 2026-06-24T12:52:45.433000+00:00

https://github.com/tscircuit/kicad-to-circuit-json/pull/155

## Abse — 2026-06-24T12:52:59.565000+00:00

bun linked to kicad mode cashe and tested it works

## Abse — 2026-06-24T13:07:50.505000+00:00

<@809856037376491570> turns out the issue is in kicadts rejects valid KiCad fp_poly nodes that lack both uuid and tstamp

## Abse — 2026-06-24T13:08:17.744000+00:00

should we make FpPoly.fromSexprPrimitives allow missing uuid/tstamp, same as KiCad itself does ?

## Abse — 2026-06-24T13:09:13.646000+00:00

[attachment]
Attachment: image.png — https://community.tscircuit.com/media/1519328547838365856

## Abse — 2026-06-24T13:09:27.790000+00:00

wdyt ?

## Abse — 2026-06-24T13:11:59.473000+00:00

but should we generate a random id or not <@809856037376491570> <@757706909351411845>

## Abse — 2026-06-24T13:12:08.464000+00:00

or just parse it without one

## Rishabh — 2026-06-24T13:15:22.896000+00:00

Yeah, kicadts should be the place where you should be fixing it. Don't generate one

## Rishabh — 2026-06-24T13:15:36.267000+00:00

The file is the source of truth, don't add anything extra

## Abse — 2026-06-24T13:16:48.404000+00:00

ok sounds good

## Abse — 2026-06-24T13:50:17.332000+00:00

<@809856037376491570> https://github.com/tscircuit/kicadts/pull/55

## Abse — 2026-06-24T14:37:39.391000+00:00

<@809856037376491570> can i get another review on this ty https://github.com/tscircuit/kicad-to-circuit-json/pull/155

## Abse — 2026-06-24T15:04:51.296000+00:00

<@809856037376491570> https://github.com/tscircuit/kicad-mod-cache/pull/27

## Abse — 2026-06-24T17:02:42.694000+00:00

<@809856037376491570> <@757706909351411845> 
I checked the difference between the old `tscircuit/kicad-component-converter` and the current `kicad-to-circuit-json` parsing for KiCad reference text.

The old converter had special handling for the font size of `property "Reference"` text. Instead of using the property’s own KiCad font size, it looked at `fp_text` entries:

- Prefer reference-like `fp_text` on `F.SilkS`
- If not found, fall back to reference-like `fp_text` on `F.Fab`
- Use that `fp_text.effects.font.size` as the font size for the emitted `pcb_silkscreen_text`

That matters for footprints like `R_0402_1005Metric`, where KiCad has:

- `property "Reference" "REF**"` on `F.SilkS` with `(size 1 1)`
- `fp_text user "${REFERENCE}"` on `F.Fab` with `(size 0.26 0.26)`

The old converter would use `0.26` from the `${REFERENCE}` `fp_text` fallback.

The current `kicad-to-circuit-json` preserves the `property "Reference"` font size directly, so it emits `pcb_silkscreen_text` with `font_size: 1`.

Then `@tscircuit/core` consumes that footprint text and adds `+0.2` via `_footprinterFontSize = elm.font_size + 0.2`, making the rendered refdes `1.2mm`. That’s why example 18’s `R1` is much larger now.

So the core behavior difference is:

Old converter:
`property Reference` font size came from small reference-like `fp_text`, often `0.26`.

New converter:
`property Reference` font size comes from the KiCad property itself, often `1`.

This is why the ref text changed size after moving to `kicad-to-circuit-json`.

## Abse — 2026-06-24T17:03:56.274000+00:00

should we change kicad to circuit json to do the same ?
Attachment: image.png — https://community.tscircuit.com/media/1519387614812307597

## Abse — 2026-06-24T17:04:18.604000+00:00

on the left is when we used kicad component converter

## Seve — 2026-06-24T17:21:09.805000+00:00

this is a nice analysis, i mean what does this look like in kicad? I think KiCad has ridiculously large fab text

## Seve — 2026-06-24T17:21:29.724000+00:00

if kicad has giant fab text, then the right side with the large fab text is correct

## Seve — 2026-06-24T17:21:51.955000+00:00

likewise, silkscreen is often much bigger than people would expect for small components. This is because silkscreen often cannot print very small

## Seve — 2026-06-24T17:22:06.754000+00:00

so ultimately the answer is however kicad displays it!!!

## Abse — 2026-06-24T18:32:21.844000+00:00

I will check kicad but I do believe big is right , u can also ask Sangit when debugging the pico small fps like 0402 silkscreen in unreadable

## Abse — 2026-06-24T18:34:34.935000+00:00

kicad is using big font
Attachment: image.png — https://community.tscircuit.com/media/1519410426444513422

## Abse — 2026-06-24T18:34:41.827000+00:00

<@757706909351411845>

## Seve — 2026-06-24T18:34:43.434000+00:00

Then thats it

## Abse — 2026-06-24T18:35:05.682000+00:00

sounds good, will see if anything needs updating