Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

lay groundwork for a possible future architecture #1571

Merged
merged 9 commits into from
Jul 31, 2022

Conversation

majbthrd
Copy link
Contributor

@majbthrd majbthrd commented Jul 28, 2022

Executive summary: lay some groundwork for adding future architectures such as the STM32WL whilst making minimum changes to the existing source code

This PR is NOT to add the STM32WL; it just tweaks the way the build is configured so that individuals can more easily add architectures other than ESP32 and NRF52.

My own general plan to add an architecture is (spare time permitting) to do a USB-only port of Meshtastic-device to the STM32WL family. The initial firmware size (about half the flash available in the part) is much smaller than the over 300kByte image of the existing NRF52 port thanks to the absence of Bluetooth, USB stack, screen driver, and other assorted extras.

#defines added are:
NO_NRF (continuing the existing NO_ESP32 convention)
NO_BUTTON (continuing the existing convention of NO_SCREEN, NO_GPS, NO_WIRE, .etc) to omit code associated with pushbuttons, rotary controls, etc.
NO_TELEMETRY (this also follows the existing convention; it replaces existing use of PORTDUINO in ./src/modules/Modules.cpp, since omission of telemetry is not specific to PORTDUINO)

Here is the nickel tour:

./platform.ini & variants/portduino/platformio.ini : exclude the future stm32wl subdirectory for existing architectures

./src/configuration.h : make the existing "HAS_WIFI" define specific to ESP32 and PORTDUINO

main.cpp & ./src/modules/Modules.cpp : introduce NO_BUTTON to omit code associated with pushbuttons, rotary controls, etc.

./src/mesh/mesh-pb-constants.cpp : the functions readcb() and writecb() are called from only one file, and that is ./src/mesh/NodeDB.cpp. The caller code is inside an #ifdef FSCom; the same #ifdef condition is applied to functions readcb() and writecb() in /src/mesh/mesh-pb-constants.cpp

./src/modules/Modules.cpp & ./variants/portduino/variant.h : replace usage of "#ifndef PORTDUINO" with an "#ifndef NO_TELEMETRY" that can be utilized by architectures other than PORTDUINO.

@github-actions
Copy link
Contributor

github-actions bot commented Jul 28, 2022

🤖 Pull request artifacts

file commit
pr1571-firmware-1.3.27.524b829.zip 524b829

thebentern added a commit to meshtastic/artifacts that referenced this pull request Jul 28, 2022
@garthvh
Copy link
Member

garthvh commented Jul 28, 2022

I would suggest maybe using functionality based flags, NO_ESP32 has been imperfect and would be better if we knew if wifi or psram was the actual functionality blocker.

@caveman99
Copy link
Sponsor Member

caveman99 commented Jul 28, 2022

The change of the Portduino/Telemetry define is fine, but please make sure the new define actually gets set when using the portduino target. Apart from that i second the remarks by @garthvh . Let's rather get rid of negative defines regarding the architecture and shift to positive defines. I have a similar changeset for RP2040 support that compiles (tm) but does not run (tm) yet. :-)

@majbthrd
Copy link
Contributor Author

The change of the Portduino/Telemetry define is fine, but please make sure the new define actually gets set when using the portduino target.

@caveman99, is there an example that you can give where this is not the case? The addition of #define NO_TELEMETRY to ./variants/portduino/variant.h was specifically intended to do just that. The existing code doesn't have a consistent convention, but at least some variants use #define NO_x in their variant.h

@majbthrd
Copy link
Contributor Author

@garthvh and @caveman99, I'd be very much in favor of a switch from the existing practice of negative "opt-out" defines to positive "opt-in" defines, and will follow up with an additional commit. That will force more source code lines to change, but I see no way to avoid this.

@caveman99
Copy link
Sponsor Member

@majbthrd please disregard my comment about the telemetry define for portduino, i missed the change at the very bottom where you included it.

thebentern added a commit to meshtastic/artifacts that referenced this pull request Jul 29, 2022
@majbthrd
Copy link
Contributor Author

majbthrd commented Jul 29, 2022

Hmm... something may not be right with the Github build system. The three targets that fail there (rak4631, rak4631_eink, and t-echo) compile locally without issue. In the Github build, it is failing on a line of code (Screen.h:11) that is #ifdef-ed out for those and other NRF52 targets. Does anyone have any clues as to what is going on?

Regardless, here's the description of the PR:

The existing not-NO_ESP32, PORTDUINO, and intermittent use of NRF52_SERIES scheme was replaced with one of these architecture definitions:

ARCH_ESP32
ARCH_NRF52
ARCH_PORTDUINO

Instead of the existing opt-out choices of NO_SCREEN, NO_GPS, NO_WIRE and USE_SIM_RADIO, and opt-in HAS_WIFI, there are a collection of purely opt-in choices:

HAS_WIFI
HAS_SCREEN
HAS_WIRE
HAS_GPS
HAS_BUTTON
HAS_TELEMETRY
HAS_RADIO
HAS_RTC

The above defines use a non-zero value to indicate an opt-in. Since there was a "HAS_EINK" in the code that could be confused with the above (since it is a choice of display rather than opting-in for a display), it was replaced with a "USE_EINK" that is more consistent with other uses in the source code of a USE_x.

Much of ./src/configuration.h and ./src/DebugConfiguration.h was moved out of those files and into one of several architecture.h files. An architecture.h unique to the architecture goes in the include directory for the architecture, i.e. ./src/esp32 or ./src/nrf52 or ./src/portduino.

As you can see in configuration.h, there is a three step process:

1: variant.h allows opt-out of HAS_ options
2: architecture.h does an opt-in of HAS_ options that variant hasn't opt-ed out of
3: configuration.h zeros all HAS_ options not touched by steps 1 and 2

I also moved the long list of #define HW_VENDOR out of configuration.h and into the corresponding architecture.h.

This hopefully makes things more modular and extensible for board variants and different target architectures.

thebentern added a commit to meshtastic/artifacts that referenced this pull request Jul 30, 2022
thebentern added a commit to meshtastic/artifacts that referenced this pull request Jul 30, 2022
thebentern added a commit to meshtastic/artifacts that referenced this pull request Jul 30, 2022
thebentern added a commit to meshtastic/artifacts that referenced this pull request Jul 30, 2022
@majbthrd
Copy link
Contributor Author

Executive summary: not all include paths in Meshtastic-device necessarily include configuration.h

Since the existing code uses configuration.h defines to opt-out of features (rather than opt-in), the earlier code could forget to include configuration.h and not notice.

@thebentern
Copy link
Contributor

Great work!
This is a big step in maintainability and supporting new hardware.
I did some basic regresssion testing on T-Beam, RAK-4631, and TLoRA v2 1.6 yesterday and everything seemed fine. So I'm going to merge. If we find anything missed, we can fix that in a future PR.

@thebentern thebentern merged commit ade32b1 into meshtastic:master Jul 31, 2022
@majbthrd majbthrd deleted the futurearch branch July 31, 2022 15:03
@geeksville
Copy link
Member

This pull request has been mentioned on Meshtastic. There might be relevant details there:

https://meshtastic.discourse.group/t/meshtastic-1-3-29-public-preview/6241/1

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Labels
None yet
Projects
None yet
Development

Successfully merging this pull request may close these issues.

None yet

5 participants