Repository navigation
RTC save writes the weekday in binary; the RX8130CE's WEEK register is one-hot #39
Description
zephcore_rtc_save() writes the day of the week as a binary 0-6 (0 = Sunday) into whichever time-block byte the date does not use. That suits the DS3231, PCF8563 and RV3028. It is wrong for the RX8130CE, whose WEEK register (13h) is one-hot.
The manual: Epson RX8130CE Application Manual ETM50E-10, §14.1.2 "Week counter", Table 15, p. 27: "The day (of the week) is indicated by 7 bits, bit 0 to bit 6", Sunday 01h, Monday 02h … Saturday 40h, and "Do not set '1' to more than one day at the same time."
What ZephCore writes:
| Day | Should be | ZephCore writes |
|---|---|---|
| Sunday | 01h |
00h: no day |
| Monday | 02h |
01h: Sunday |
| Wednesday | 08h |
03h: two days at once |
| Saturday | 40h |
06h: two days at once |
The time and date are unaffected, since ZephCore never reads the weekday back, but the chip's week counter holds an invalid value after every sync. It matters only to whatever reads it, such as a weekday alarm or another firmware.
Affected boards: any that include rtc-i2c.dtsi with an RX8130CE fitted at 0x32. meshtracker_x1's own 0x32 descriptor (RX8900 layout) needs the same check against that part's data sheet.
Fix direction: a per-descriptor property for the weekday encoding (binary or one-hot), set on the RX8130CE descriptor, with the save writing 1 << dow for one-hot.
Found by the local review of #38. Not fixed there, being pre-existing and outside its scope.
Activity
ptr727 commentedon Oct 4, 2026
MeshCore has the same defect:
RTC_RX8130CE::setTime()insrc/helpers/RTC_RX8130CE.cppwritesbin2bcd(t->tm_wday) & 0x07to 13h, andgetTime()reads it back the same way (checked at ptr727/meshcore-dev-MeshCoredev3e3150c, which includes the latest upstream merge). Passed to the MeshCore session to decide its upstream handling.ptr727 commentedon Oct 4, 2026
MeshCore's fix for the same defect is upstream: meshcore-dev/MeshCore#3550 (issue meshcore-dev/MeshCore#3549), writing
1 << tm_wdayand decoding the lowest set bit. ZephCore's fix should follow the same encoding.ptr727 commentedon Oct 5, 2026
A lead on the open meshtracker_x1 part, from the local review of #45 after
dev(liquidraver#98) was merged in. Not proven.rtc-ysn8900@32node uses the RX8900 register layout, and its weekday is still written as 0-6.epson,rx8900throughrtc-rv8803.c, which writes the weekday as1 << tm_wday. If the YSN8900 follows the RX8900 there, this board has the same wrong-day defect as the RX8130CE, and the fix is to addweekday-one-hot;to that node.ptr727 commentedon Oct 5, 2026
Resolved for meshtracker_x1. The Epson RX8900 Application Manual (ETM45E-11, §8.2.2, Table 15, p. 16) gives WEEK (03h) as one-hot, Sunday
01hto Saturday40h. The YSN8900 brief has no register map, but its pinout matches the RX8900CE pin for pin, and the board already uses the RX8900 map. With Pieter's approval,weekday-one-hotis now set on thertc-ysn8900@32node and folded into liquidraver#101 (head 2621782). The YSN8900 programming guide is still unavailable, so this rests on it being an RX8900 second source. MeshCore is unaffected: its meshtracker_x1 never drives the chip.ptr727 commentedon Oct 10, 2026
Fixed upstream by liquidraver#101, merged 2026-10-06: the RX8130CE and YSN8900 weekday is written one-hot.