ESP-32 build error: no member named 'comd'

[hayschan]

Environment ( 実行環境 )

• MCU or Board name: ESP32-S3
• Panel Driver IC: ILI9488
• Bus type: Parallel
• LovyanGFX version: 1.1.8 (commit ID: 19a2e71)
• Build Environment : VSCode ESP-IDF
• Operating System: Windows 11 Pro

Hello. I'm reporting because I think I found a bug.

I think this may be related to the fix published about the ESP-IDF error mentioned here. It might cause a new bug. #423

I'm using ESP32-S3 and I wrote the code with ESP-IDF v5.0.2.

Symptom

The error codes are as follows.

./components/LovyanGFX/src/lgfx/v1/platforms/esp32/common.cpp: In function 'void lgfx::v1::i2c::i2c_set_cmd(i2c_dev_t*, uint8_t, uint8_t, uint8_t, bool)':

./components/LovyanGFX/src/lgfx/v1/platforms/esp32/common.cpp:808:14: error: 'struct i2c_dev_t' has no member named 'comd'; did you mean 'comd0'?
  808 |       (&dev->comd[0])[index].val = cmd_val;
      |              ^~~~
      |              comd0

My fix

When I change the line of code from (&dev->comd[0]) to (&dev->comd0), the build errors disappear.

Before:

// common.cpp:806
#if defined (CONFIG_IDF_TARGET_ESP32S3)
 #if (ESP_IDF_VERSION >= ESP_IDF_VERSION_VAL(5, 0, 2) && ESP_IDF_VERSION < ESP_IDF_VERSION_VAL(5, 1, 0))
      (&dev->comd[0])[index].val = cmd_val;
 #else
      (&dev->comd0)[index].val = cmd_val;
 #endif
#else

After:

// common.cpp:806
#if defined (CONFIG_IDF_TARGET_ESP32S3)
 #if (ESP_IDF_VERSION >= ESP_IDF_VERSION_VAL(5, 0, 2) && ESP_IDF_VERSION < ESP_IDF_VERSION_VAL(5, 1, 0))
      (&dev->comd0)[index].val = cmd_val;
 #else
      (&dev->comd0)[index].val = cmd_val;
 #endif
#else

[lovyan03]

Thanks for reporting back.
It is necessary to identify exactly from which version of the IDF the notation has changed.

[hayschan]

My version is v5.0.2.

> idf.py --version
ESP-IDF v5.0.2-dirty

[lovyan03]

Basically, only the released version is supported.
I cannot guarantee the operation of other versions.

[hayschan]

• In v5.0.3, it is comd.
• In v5.0.2, it is comd0.

(Fuc* ESP) 😠

I will track down every major version, and help with creating a detailed #if #else statements regarding every major release version. I'll fork, contribute and pull request soon. Let me know if you think this will work.

v5.0.3

    volatile i2c_comd_reg_t comd[8];
    // ...
} i2c_dev_t;

Link to code: https://github.com/espressif/esp-idf/blob/v5.0.3/components/soc/esp32s3/include/soc/i2c_struct.h#L996

v5.0.2

    volatile i2c_comd0_reg_t comd0;
    // ...
} i2c_dev_t;

Link to code: https://github.com/espressif/esp-idf/blob/v5.0.2/components/soc/esp32s3/include/soc/i2c_struct.h#L1161

[tobozo]

@hayschan can you try with this macro?

#if defined (CONFIG_IDF_TARGET_ESP32S3)
 #if (ESP_IDF_VERSION > ESP_IDF_VERSION_VAL(5, 0, 2) && ESP_IDF_VERSION < ESP_IDF_VERSION_VAL(5, 1, 0))
      (&dev->comd[0])[index].val = cmd_val;
 #else
      (&dev->comd0)[index].val = cmd_val;
 #endif
#else

[hayschan]

Yep, it works, as I would be running the #else statement part.

[hayschan]

comd, comd[] in every major ESP-IDF version

I just checked ESP-IDF every major version commit tags. In v5.1, it changes back to comd0 again. 😕 Only a single ESP-IDF version v5.0.3 uses comd[8].

It seems that comd[8] comes out of no where.

Version	Variable type	Link to reference (to the exact line)
v5.1	comd0	esp-idf/components/soc/esp32s3/include/soc/i2c_struct.h at v5.1 · espressif/esp-idf (github.com)
v5.0.3	comd[8]	esp-idf/components/soc/esp32s3/include/soc/i2c_struct.h at v5.0.3 · espressif/esp-idf (github.com)
v5.0.2	comd0	esp-idf/components/soc/esp32s3/include/soc/i2c_struct.h at v5.0.2 · espressif/esp-idf (github.com)
v4.4.5	comd0	esp-idf/components/soc/esp32s3/include/soc/i2c_struct.h at v4.4.5 · espressif/esp-idf (github.com)
v4.3.5	N/A, does not exist this variable	esp-idf/components/soc/esp32s3/include/soc/i2c_struct.h at v4.3.5 · espressif/esp-idf (github.com)

Possible fix suggestion

As only a single ESP-IDF version v5.0.3 changes to comd[8], we could just make a single #if targeting this version.

#if defined (CONFIG_IDF_TARGET_ESP32S3)
 #if (ESP_IDF_VERSION == ESP_IDF_VERSION_VAL(5, 0, 3))
      (&dev->comd[0])[index].val = cmd_val;
 #else
      (&dev->comd0)[index].val = cmd_val;
 #endif
#else

[tobozo]

esp-idf sure provides a lot of efforts to scare away maintainers ... hopefully their inclusion in the Matter project will break a lot of eggs with backporting and bring instant karma to whomever dictated such a schizophrenic strategy

[lovyan03]

I would like to consider the following possibilities

• Possibility that 5.0.4 is created as a successor version to 5.0.3
• Possibility that 5.1.1 is created as a successor version of 5.1 and cond[0] is used.

Perhaps it might be faster to access the register using the address definition instead of using the ESP-IDF structure definition to solve the problem.

[smeisner]

Using the latest (as far as I can tell) ESP IDF (v5.1.1) on an ESP32-S3, the only way I could get this to work was to change the second test to be less than 5.1.2 ... as follows:

// common.cpp:806
#if defined (CONFIG_IDF_TARGET_ESP32S3)
 #if (ESP_IDF_VERSION >= ESP_IDF_VERSION_VAL(5, 0, 2) && ESP_IDF_VERSION < ESP_IDF_VERSION_VAL(5, 1, 2))
      (&dev->comd0)[index].val = cmd_val;
 #else
      (&dev->comd0)[index].val = cmd_val;
 #endif

[lovyan03]

@smeisner
The content after the conditional branching seems to be exactly the same?

[smeisner]

@lovyan03 Notice I changed the test from "5, 1, 0" to "5, 1, 2"