PCF8563's second/minute/hour/day/weekday/month/year registers are
contiguous (0x02-0x08), so PCF8563_ReadDateTime() bursts all 7 bytes
in one transaction. PCF8563_ReadTime() becomes a thin wrapper that
discards the date fields. WriteTime() is kept as its own 3-byte-only
implementation (writing only sec/min/hour) rather than routed through
the new WriteDateTime(), since forcing it through a 7-byte write would
require synthesizing date values and risk corrupting the stored date;
WriteDateTime() is added as a separate full 7-byte write for callers
that actually have real date+weekday values to persist.
SGM_CTRL was turned on at the very start of LCD_Init(), before the
ST7789V register sequence and first frame ran, showing garbage/blank
content briefly on every boot. GPIO_Init() now forces backlight off at
power-on, LCD_Init() no longer touches it, and main() only turns it on
right before entering the main loop (after HomeApp's first frame is
drawn). Wakeup_Restore()'s existing on-then-reinit sequence is
untouched since it sets the level itself before calling LCD_Init().
Root cause found via datasheet: PxIM0/PxIM1 has no dual-edge mode, only
falling/rising/low-level/high-level. (IM1=1,IM0=1) was actually
high-level interrupt, refiring continuously for the whole high-pulse
duration (confirmed by scope: no real HF noise, and edge_count=0 when
P3.6 grounded). Fix: start in rising-edge mode, software-toggle to the
opposite edge on every trigger (ping-pong) to emulate true dual-edge
triggering, feeding exact edge direction into the incremental decode
state machine (no glitch filter needed). Also switch P3.6 idle bias
from pull-up to pull-down, since floating-high under the old high-level
mode caused the interrupt-storm boot issue; removed the now-conflicting
pull-up restore in Wakeup_Restore().
Diagnostic experiment: restore the commit d8df5d4 polling decode
algorithm/thresholds unchanged, while P3.6 ISR now only increments an
edge counter (no decode state machine). Successful decodes print the
edge-count delta over the decode window to compare against the
EV1527 theoretical 50-edges-per-frame, to validate/refute the
edge-interrupt glitch-noise hypothesis.
TH0/TL0 torn-read protection didn't cover the case where the hardware
counter already wrapped but Timer0_Isr hasn't run yet to bump
timer0_ovf_count, making the timestamp appear to jump backward by one
overflow period (~32.768ms). Read TF0 under a brief EA guard and treat
a pending-but-unserviced overflow as ovf+1 without clearing it.
