Add ESP-IDF 6.0 support with dual 5.x/6.x compatibility

[Bwooce]

Summary

Adds ESP-IDF 6.0 support while maintaining full backward compatibility with IDF 5.x (tested against 5.5). All changes use #if ESP_IDF_VERSION guards so both major versions compile cleanly for both ESP32 and ESP32-S3 targets.

Closes #450

Detailed changes

Build system (CMakeLists.txt)

• Added IDF 6.x branch for component registration: IDF 6 splits legacy driver into sub-components (esp_hal_dma, esp_driver_rmt, esp_driver_gpio), so these are now listed as explicit REQUIRES for IDF_VERSION_MAJOR > 5
• Existing IDF 4.x and 5.x branches are untouched

LCD driver — ESP32-S3 (src/output_lcd/lcd_driver.c)

• LCD peripheral signals: struct renamed from lcd_periph_signals.panels[0].member / lcd_periph_rgb_signals.panels[0].member to array soc_lcd_rgb_signals[0].member in IDF 6. Replaced LCD_PERIPH_SIGNALS macro with LCD_PERIPH_SIG(member) to abstract the access pattern across all three naming conventions
• LCD signal struct: forward-declared soc_lcd_rgb_signal_desc_t and extern soc_lcd_rgb_signals directly instead of including hal/lcd_periph.h, which in IDF 6 transitively pulls in hal/i2s_ll.h causing conflicts
• GDMA API: gdma_new_channel() replaced by gdma_new_ahb_channel() in IDF 6 — takes 3 args (config, tx_chan, rx_chan) and gdma_channel_alloc_config_t no longer has a direction member
• RMT peripheral clock: periph_module_enable(PERIPH_RMT_MODULE) removed in IDF 6 (PERIPH_RMT_MODULE no longer exists in shared_periph_module_t). Replaced with rmt_ll_enable_bus_clock() + rmt_ll_reset_register() inside PERIPH_RCC_ATOMIC() blocks
• RMT LL rename: rmt_ll_enable_periph_clock() renamed to rmt_ll_enable_group_clock() in IDF 6
• RMT types: rmt_item32_t removed with legacy RMT driver in IDF 6 — defined locally with identical layout
• RMT signals struct: rmt_periph_signals.groups[0] renamed to array soc_rmt_signals[0] in IDF 6
• RMT channel count: SOC_RMT_CHANNELS_PER_GROUP removed — replaced with RMT_LL_CHANS_PER_INST from hal/rmt_ll.h
• RMT periph header: moved from soc/rmt_periph.h to hal/rmt_periph.h in IDF 6
• LCD LL interrupt macro: IDF 5.5+ added a wrapper macro around lcd_ll_enable_interrupt() that references __DECLARE_RCC_ATOMIC_ENV, requiring the caller to be inside a PERIPH_RCC_ATOMIC() block. Guard lowered to >= 5.5.0 (not just 6.0)

RMT pulse driver — ESP32 (src/output_i2s/rmt_pulse.c)

• Legacy RMT driver removed: driver/rmt.h (rmt_config(), rmt_set_tx_intr_en(), rmt_config_t) completely removed in IDF 6. Replaced init with direct register writes matching what rmt_config() did internally — the actual pulse generation already used direct RMTMEM access
• RMT types: locally defined rmt_item32_t and rmt_block_mem_t for IDF 6 (same approach as lcd_driver.c)
• Peripheral clock: same PERIPH_RCC_ATOMIC() + LL function pattern as lcd_driver.c
• GPIO routing: gpio_matrix_out() replaced with esp_rom_gpio_connect_out_signal() via esp_rom_gpio.h

I2S data bus — ESP32 (src/output_i2s/i2s_data_bus.c)

• Peripheral clock: periph_module_enable(PERIPH_I2S1_MODULE) removed in IDF 6 — replaced with i2s_ll_enable_bus_clock(1, true) + i2s_ll_reset_register(1) inside PERIPH_RCC_ATOMIC(). Same for disable in i2s_bus_deinit()
• GPIO signal macros: I2S1O_DATA_OUT0_IDX / I2S1O_WS_OUT_IDX lost their transitive include path in IDF 6 — added explicit #include "soc/gpio_sig_map.h"
• GPIO routing: gpio_matrix_out() replaced with esp_rom_gpio_connect_out_signal() for IDF 6

ADC temperature sensing (src/board/epd_board_common.c)

• Legacy ADC API removed: adc1_get_raw(), esp_adc_cal_characterize(), esp_adc_cal_raw_to_voltage() all removed in IDF 6
• Reimplemented epd_board_temperature_init_v2() and epd_board_ambient_temperature_v2() using the oneshot ADC driver (adc_oneshot_new_unit, adc_oneshot_config_channel, adc_oneshot_read) + new calibration API (adc_cali_create_scheme_line_fitting, adc_cali_raw_to_voltage)

Cache header path (src/output_lcd/render_lcd.c)

• rom/cache.h may be relocated in IDF 6 — added __has_include fallback to esp32s3/rom/cache.h

SPIRAM config (src/highlevel.c)

• IDF 6 unifies CONFIG_ESP32_SPIRAM_SUPPORT / CONFIG_ESP32S3_SPIRAM_SUPPORT under CONFIG_SPIRAM — added it as an additional check

Demo build (examples/demo/main/CMakeLists.txt)

• GCC 15 (shipped with IDF 6) treats -Wbidi-chars= as an error for Unicode bidirectional control characters in the generated font header files. Added -Wno-error=bidi-chars= compile option

CI (.github/workflows/main.yml)

• Added release-v6.0 to the build matrix for both ESP32 and ESP32-S3 (switch to v6.0 once the stable Docker tag exists)
• Added IDF 6 extended examples (mirrors the v5.4 set)
• Set continue-on-error: true for IDF 6 while it is pre-release
• IDF 6 kconfgen crashes on large sdkconfig.defaults from IDF 5 (upstream bug: choice.selection is None). CI uses minimal sdkconfig.defaults for IDF 6 builds as a workaround
• Added comment noting Arduino-ESP32 IDF 6 support is pending upstream

CI results

Target	IDF 5.2	IDF 5.3	IDF 5.4	IDF 6.0
ESP32-S3 demo	PASS	PASS	PASS	PASS
ESP32 demo	PASS	PASS	PASS	PASS
screen_diag	—	—	PASS	PASS
dragon	—	—	PASS	PASS
grayscale_test	—	—	PASS	PASS
calibration_helper	—	—	PASS	PASS
www-image	—	—	PASS	FAIL (*)
Arduino weather	PASS (v5.3.2)

(*) www-image fails on IDF 6 due to GCC 15 warnings-as-errors in the third-party bitbank2/jpegdec managed component (-Werror=pointer-to-int-cast, -Werror=shift-negative-value), not epdiy code. Requires an upstream fix to jpegdec.

Test plan

• CI passes for IDF 5.2, 5.3, 5.4 (all green)
• CI passes for IDF 6.0 ESP32 and ESP32-S3 demo builds
• CI passes for IDF 6.0 extended examples (except www-image — third-party dependency)
• Verify demo runs correctly on physical ESP32-S3 hardware with IDF 6.0
• Verify demo runs correctly on physical ESP32 hardware with IDF 6.0
• Once IDF 6.0 reaches stable release, switch release-v6.0 to v6.0 and remove continue-on-error

[Bwooce]

(yes, created with Claude the clanker. Hence the way longer description than I would have written, but it's useful to see the reasoning). I'll get the tests passing before clicking ready-for-review,

[martinberlin]

@Bwooce please fix as soon as possible the failed automatic checks if you want to get this merged

[Bwooce]

All checks are now passing. The one remaining failure is www-image on IDF 6.0, which is caused by GCC 15 warnings-as-errors in the third-party bitbank2/jpegdec managed component (shift-negative-value, pointer-to-int-cast), not epdiy code. I've submitted a fix upstream (bitbank2/JPEGDEC#113), waiting on their review.

The IDF 6.0 jobs have continue-on-error: true so the overall workflow reports success, but the individual check still shows as failed.

[martinberlin]

@Bwooce what is the status here?
Shall I write Larry Bank and ask about an update in JPEGDEC?

I will try to add your PR about I2C next weekend. Sorry but it was stale, and not working properly, in the tests I run.

[Bwooce]

I'll go chase the missing upstream PR Martin, it's not a very active repo. I could also try and work around it, it's literally due to a "warnings now treated as errors" change in IDF 6 which uses GCC 15 so the PR is nice-to-have functionally but working around it long-term would be incorrect.

[martinberlin]

No problem. I will try to test your update. I've a v7 board to share with you if you would like to try it, just need to pay the shipping.

[vroland]

Hi, sorry to be late to the party, just a suggestion if it doesn't burn too many tokens: We could separate the ADC and RMT migrations into separate PRs, since those should be backwards compatible. The ADC I can test while travelling, the RMT I can't test right now because I don't have an older board with me.
Then we can rebase the 6.0 branch, which should become much smaller. At that point, if the JPEGDEC dependency is not updated, I'd be fine with moving the www-image example into a separate repo that only builds with 5.x or hacks, while all the remaining examples should build cleanly with 6.0.

[vroland]

I think this may not need to be conditional, since the oneshot adc driver was introduced in IDF V5, and we don't support IDF 4.x anymore. Then we remove the old implementation. Could also be moved into a separate PR.

[vroland]

Same here, I don't think we gain much by keeping both implementations. We can just migrate to the new implementation. Also maybe in a separate PR?

[vroland]

This would be included in the rmt migration, then it doesn't need to be duplicated in I2S and LCD driver.

[martinberlin]

@vroland in the top of that further upgrades of IDF won't let you use the old I2C legacy driver. This happened already with me on a C5 board (Yes C5 works also to drive parallel using the new Parlio driver)

https://github.com/martinberlin/i2c_bus this is the super simple component I built so you can just use one I2C bus for many different I2C peripherals without running into the warning "I2C is already initialised" this would be also something cool to implement into epdiy. As many people uses it for home automation, connecting other i2c sensors to the board.

[vroland]

If I understand your component correctly, it would still require people to use your component if they want to re-use the I2C bus? If that's the case, wouldn't it make more sense to migrate to the new I2C driver and provide a way to optionally initialize epdiy with an existing bus? Then we don't introduce another custom way of doing things.

[martinberlin]

If I understand your component correctly, it would still require people to use your component if they want to re-use the I2C bus? If that's the case, wouldn't it make more sense to migrate to the new I2C driver and provide a way to optionally initialize epdiy with an existing bus? Then we don't introduce another custom way of doing things.

Yeah sure what I was saying is that it would be nice to use something like this inside epdiy itself. So then we have only one file to modify where I2C is handled making it easy to update it without searching in what other parts of the code is instantiated (PCA9535, TPS65185)
Just an idea that I though can bring a more cohesive and maintainable code

[vroland]

Superseeded by #477