Homepoint 2.0 Announcement

[sieren]

I am currently working on Homepoint 2.0 which will be a complete rewrite based on https://github.com/lvgl/lvgl, allowing to more easily support new platforms and a more responsive UI.

[sieren]

Beta versions for AZTouchMod and M5Stack Core 2

homepoint_m5core2_aztouchmod_betas.zip

[Girgitt]

Initial impressions on M5Core2 are very good. The lvgl-based version seems stable (at least no unexpected reboots experienced).

However with the screen layout copied from the example config the initial screen is gray with IP address appearing after a couple of seconds. To get proper screen content one must tap to change screen and then move back manually triggering redraw/refresh.

Do you have plans to continue development?

Your approach to navigation seems better compared to the much simpler code of cfGUI library which I extended to support M5core2 but any other development is probably worthless given how functional and eye-candy your design is.

EDIT: Actually the initial gray screen does not always happen. On some occasions reset results in a properly drawn initial screen. Maybe the issue results from the lack of topics on my test MQTT server.

[sieren]

Thanks for your assessment @Girgitt :) There are still tons of bugs in there and the code is still somewhat messy as I've been trying to do performance improvements to increase FPS.

It's true, I haven't contributed much to the repo recently, but I have not abandoned the project. It's just that I had very little time to work on side projects in the past 6 months but I hope to return sometime soon.

[Girgitt]

That's some good news :) Thank you for reaching back. I wish my Cpp skills were much better to be able to provide some valuable contribution but who knows, maybe it's a good enough reason to at least try when M5core2 becomes the supplemental HMI to my Homematic IP network in the next couple of months.

[Girgitt]

Hi @sieren would find time to update/add building steps for homepoint-2.0-lvgl branch?

Instructions from wiki are obsolete as esp-idf version changed to v4.4 (at least this is what I found based on arduino-esp32 submodule's history for version committed to the homepoint-2.0-lvgl branch) but even with an updated esp-idf the compilation using /scripts/build_all.sh fails due to error: homepoint/components/ESPAsyncWebServer/src/WebAuthentication.cpp:74: undefined reference to `mbedtls_md5_starts'

Maybe some specific tag of esp-idf should be used?

(There are other issues like code from /scripts/merge_bin_esp.py being written in py2 while esp-idf uses py3 with virtualenv which makes the environment difficult to initialize. Fortunately, a trivial fix is to call python2 in build_all.sh instead of python),

EDIT: with esp-idf v4.4.1 the mentioned error is gone. M5Stack2 boots but shows error: "could not pares failsafe file".
The homepoint.bin fails to build for targets M5Stack and ESP32 without any specific error - just file not found exception is thrown by merge_bin_esp.py. `

[greiner11]

I concur with @Girgitt opinion. I am also enthusiastic about the project - it's a pity that it has not been developed further for some time. It would also help me if I could compile it myself. However, I also failed with several errors. The branch from Homepoint 2.0 has a VSC-Workspace. Is it possible to compile the code with Visual Code and the extension espressif (esp-idf extension)?

[Girgitt]

So I tried to re-visit the project after a year and OMG what an adventure. First of all, despite what I wrote earlier, it is impossible that esp-idf version required to compile it is v4.4.1. IT IS NOT. it is something between 4.4.0 and 4.4.1. The proper version should be the esp-idf master at the time of arduino-esp32. Good luck in finding that - there is a version in submodule components/arduino-esp32 but it is Sieren's fork. Anyway by trying to use esp-idf v.4.4.1 and following error about missing reference to mbedtls_md5_starts i found commit that defines mbedtls_md5_starts_ret:

zbig@TX1320-1:~/repos/priv/esp/esp-idf_4_4_1$ git log -S "define mbedtls_md5_starts"
commit e87c64dc4443d973b784c9674a79259493abf9c0
Author: liuhan <[email protected]>
Date:   Tue Apr 6 14:21:28 2021 +0800

    mbedtls: Add CONFIG_MBEDTLS_ROM_MD5 to use ROM MD5 in mbedTLS
zbig@TX1320-1:~/repos/priv/esp/esp-idf_4_4_1$ git show e87c64dc4443d973b784c9674a79259493abf9c0 | grep esp_md5_init
+int esp_md5_init_ret( mbedtls_md5_context *ctx );
+void esp_md5_init( mbedtls_md5_context *ctx );
+#define mbedtls_md5_init                        esp_md5_init
+#define mbedtls_md5_starts_ret                  esp_md5_init_ret
+int esp_md5_init_ret( mbedtls_md5_context *ctx )
+void esp_md5_init( mbedtls_md5_context *ctx )
+    esp_md5_init_ret(ctx);

One commit earlier 1f54d17503ff17d830de75eed4eb47bc6a3aafc3 was a good candidate and it works - at least Homepoint from branch homepoint-2.0-lvgl compiles. To avoid issues one needs to: esp-idf# rm -rf ~/.esspressif esp-idf# git checkout 1f54d17503ff17d830de75eed4eb47bc6a3aafc3 esp-idf# git submodule update --init --recursive (make sure to install python3 requirements fromvim install guide: https://docs.espressif.com/projects/esp-idf/en/latest/esp32/get-started/linux-macos-setup.html ) edit ./tools/idf_tools.py line 13385 and remove --user arg to pip; at least on py3.10 this breaks install.sh esp-idf# ./install.sh esp-idf# . ./source.sh

Then (after setting up python2 in your environement and modifying build_all.sh to use python2 instead of python and optionally commenting out steps for platforms you don't need) compile Homepoint with homepoint/scripts/# ./build_all.sh

At this point I am not sure if the resulting build works but size seems correct. Time to look for USB A to C cable.. and build loaded with sudo esptool -b 921600 write_flash -fs 4MB -fm dout 0x0 ./homepoint_m5stackcore2_full.bin doesn't work (black screen). serial debug shows error: E (1214) i2c: i2c_master_cmd_begin(1394): i2c driver not installed ESP_ERROR_CHECK_WITHOUT_ABORT failed: esp_err_t 0x103 (ESP_ERR_INVALID_STATE) at 0x4008991b

Before that I tried many esp-idf versions and here is other way that resulted in compilation but M5Core2 did not start after flashing with esptool:

1. checkout esp-idf to v.4.4-beta1
2. change Homepoint/components/ESPAsyncWebServer/src/WebAuthentication.cpp:74 mbedtls_md5_starts to mbedtls_md5_starts_ret
3. add values from https://github.com/espressif/arduino-esp32/issues/5925#issuecomment-981112887 to esp-idf/components/esp_rom/esp32/ld/esp32.rom.ld

[EDIT] Tried different approach - Sieren's PR (https://github.com/espressif/arduino-esp32/commit/e30b821be13a5543fd8dfa30a00b3ea42d74107b) was merged to esspressif/arduino-esp32 and relesed in all v2.0.X tags. test 1: arduino-esp32 v2.0.0 + esp-idf v4.4 = compilation successful but initialization fails with E (1214) i2c: i2c_master_cmd_begin(1394): i2c driver not installed

• arduino-esp32 v2.0.0 cannot work - it does not include https://github.com/espressif/arduino-esp32/commit/794de50bf4819f1d47fc7090093a4e2f3097dafd Indeed I forgot to revert manual changes to linker file (https://github.com/espressif/arduino-esp32/issues/5925#issuecomment-981112887) and after reverting change compilation fails with Homepoint/components/arduino-esp32/cores/esp32/MD5Builder.cpp:97: undefined reference to `MD5Final'

the lowest possible arduino-esp32 version to use: v.2.0.2

test 2: arduino-esp32 v2.0.2 + esp-idf v4.4.1 = E (1214) i2c: i2c_master_cmd_begin(1398): i2c driver not installed ESP_ERROR_CHECK_WITHOUT_ABORT failed: esp_err_t 0x103 (ESP_ERR_INVALID_STATE) at 0x4008998b file: "./components/esp_i2c_helper/i2c_helper.c" line 100

another option is to take idf version from debugger output of the working beta build by Sieren: I (667) cpu_start: ESP-IDF: v4.0.2-206-g5630b17e1-dirty however this idf version is not compatible with commited version of arduino-esp32 submodule

build arduino-esp32 v2.0.2 + esp-idf v4.4.1 is promissing but 3 issues exist:

1. i2c driver is not loaded; it can be loaded manually in main.c:app_main and seems to work as voltages are logged in debug output
2. SPI MOSI pin (19) is different than in debugging output of beta image (38) logged by lvgl_helpers; not sure if the debug output with pin numbers used is for touch or display
3. SPIFFS fails to mount with error -10025 but build from master shows the same error and icons are displayed FIX: in lvgl branch application binary size increased but spiffs offset was not updated in /scripts/build_all.sh script in arguments passed to merge_bin_esp.py. SPIFFS offset should be 0x310000 and not 0x2b0000: --bin_address 0x8000 0xd000 0x1000 0x10000 0x310000