Skip to content

1.2.35

Choose a tag to compare

@lovyan03 lovyan03 released this 25 Oct 01:40
· 138 commits to master since this release
1b87851

Fixes

  • Panel_M5HDMI: a pushImage() sent right after a long copyRect() could overflow the FPGA's receive buffer, after which the panel answered busy until it was power-cycled. The command sent after each busy wait is now counted, so the next image waits for the copy to finish. (#979)
  • I2C on ESP32-S3 / C2 / C3 / C5 / C6 / C61 / H2 / P4: below about 78 kHz the START / STOP timing overflowed its 9 bit register field and wrapped to a much shorter time (75 kHz: START hold 0.55 µs), so devices stopped answering; on ESP32-C3 a setting below about 133 kHz also ran the clock faster than requested (100 kHz ran at about 150 kHz). Each timing value now saturates at its field width; the slowest bus on these chips is about 39 kHz. ESP32 / ESP32-S2 and settings that already fitted are unchanged. (#978, #980)
  • Panel_M5HDMI: updated the FPGA bitstream to a rebuild of the design's Veryl port. COPYRECT no longer hangs busy for a short horizontal shift of an overlapping area, a host that pauses mid stream no longer shifts the picture, and the display no longer stays busy after a transfer whose last pixel arrives while the previous write is still waiting (e.g. a 1 px wide pushImage()). The flash image grows by about 20 KB. (#981)
  • GDEW0154D67 / SSD1677 / TM1680: after init_without_reset(false), the image on the panel was wiped during initialization (the init re-applied the invert setting and the panel treated it as a full-frame change; on the TM1680 the cleared buffer was pushed). Re-applying the current value no longer refreshes the panel, and the first drawing on the GDEW0154D67 after such a start is no longer inverted. Panel_1bitOLED / Panel_M5UnitGlass clear their host buffer in init(). (#983)
  • RP2040 / RP2350 SPI: readData() / readBytes() (and readPixels()) wrote ahead of the reads without a limit, so a delay of a few byte times (an interrupt, the WiFi driver) overflowed the 8-entry RX FIFO, a byte was lost and the read hung. No more than 8 bytes are now in flight, as in the Pico SDK's spi_read_blocking(). (by @ctorkildson-netizen, #982)