Mark, Michael
Heads up: the vwegolf module merged for 3.3.006 has a CAN bus bug. AC charging triggers a
TWDT reset loop, module reboots every few seconds while plugged in.
Background: during capture analysis I saw KCAN-typical IDs (BMS SoC, gear) showing up on
FCAN, and concluded the J533 gateway was merging KCAN traffic onto FCAN. So IncomingFrameCan2 blanket-forwards every FCAN frame into the KCAN handler. The right read
is: J533 bridges a small number of IDs selectively (e.g. 0x131 SoC during DC charging when
the BMS goes quiet on KCAN), not wholesale traffic merging. Wrong assumption, wrong fix.
That alone wouldn't kill anything, but it exposed a latent mcp2515 bug where the acceptance
filter doesn't persist past Start(), so the ISR takes ~860 fps of FCAN instead of the 4-5
IDs we actually decode. Both bugs together = watchdog reset loop on AC charge.
Two fixes in flight (drafts):
- #1403 — mcp2515: persist acceptance filter across power cycles (framework-level)
- #1404 — vwegolf: drop the forwarding, install proper hardware filter, decode FCAN signals
where they belong (depends on #1403)
Mark already commented on #1403 with a cleaner approach — refactor incoming once design is
confirmed. Hardware-verified on my module.
Up to you whether to hold main, drop vwegolf from the tag, or ship + fix in 3.3.007. Sorry
for the trouble.
Sincerely,