Hi Jamie,

I think those questions were directed to Mayank, who is experiencing the issue now.

Thanks,

Carles

Hi All,

For today’s meeting let’s follow up on last meeting action items and get a status update.

Also I think we should consider the short / long term goals of this WG and who has the skills / capacity to do what.

Where:
https://zoom.us/j/967549258

Agenda

• Action items from last meeting:
• Wayne: PR22668: Continue work, and use as base for better toolchain abstraction: Status ?
• Thomas: IAR: Get basic hello world compiling, (linking with GCC) : Status ?
• Everyone: Give feedback on PR#22688, and issue: #5517: Status ?
• Toolchain WG Goals
  
  

Also some feedback has been given by Marc since last meeting:

• More reliable than code reviews, I found that gcc/clang -std=cXX -pedantic seem to work great for standard compliance. In our internal project I configured an extra build configuration with -pedantic on the side, we use it like "checkpatch" and others.

Note: Should be coupled with only checking new / modified files in a PR.

Feel free to send a mail, if you would like additional topics to be discussed.

Best regards

Torsten T. Rasmussen           

Live meeting minutes: https://docs.google.com/document/d/1IQKBK-GcJNZG0O9QArqYfvb6Huk5xHscN-XIGEZr-z8/edit#heading=h.x36xe8bnwr9r

Hi Vinayak,

> Are you using any host debugger firmware to convert to USB? or using 3v TTL connections to your custom board from nRF52840?

When I tested it, it was using an nRF52840 PCA10059-like board without a 32KHz crystal (RC was selected from zephyr), I also tried on a BL654 DVK with the 32KHz connected and set in zephyr

> For those using USB_UART conversion, I am assuming you are using Segger JLink firmware on the host chip in the nRF52840 DK/dongle. Could you try using the MBED firmware on the host debugger chip?

When I was testing the UART one, I used an FTDI chip so this does not apply. I tested against a BT850 module, which also uses an FTDI with a non-zephyr based HCI module and that worked flawlessly

> Also, please create or refer me to GH issue, so that further discussions can be continued there.

I never created an issue. I haven’t tested in probably about a year now or so.
Thanks,

Jamie

Hi Mayank,

Are you using any host debugger firmware to convert to USB? or using 3v TTL connections to your custom board from nRF52840?

For those using USB_UART conversion, I am assuming you are using Segger JLink firmware on the host chip in the nRF52840 DK/dongle. Could you try using the MBED firmware on the host debugger chip?

Also, please create or refer me to GH issue, so that further discussions can be continued there.

Regards,

Vinayak

Hi,

I’ve previously done testing with the HCI UART and USB sample apps (in older Zephyr versions) with an nRF52840 and linux PC and always found it to not work well. It would start fine, could scan for devices from bluetoothctl and then it just seemed to screw up every time with command timeouts. Never did much digging into why.

Thanks,

Jamie

> Hi All,

>

> I'm facing issues regarding ble's hci interface.

> I'm having nrf52840 module on which i run zephyr's hci_uart app for beacon scanning.

> I have Bluez 5.50 installed on my custom board (imx6ull).

>

> hci0 interface on my device is "UP RUNNING". So, i'm able to scan the beacons. (My application scans beacons after every 10 seconds).

>

> Now, When i have put my device for over night testing, then i have observed that after few hours beacon scanning was stopped.

> And i'm getting "Bluetooth: hci0 command 0x200b tx timeout" prints after every 5-10 seconds on the console, still the hci0 interface status is "UP RUNNING", but not able to scan the beacons.

>

> Sometimes also getting the prints like "Bluetooth: hci0: Frame reassembly failed (-84)".

>

> What could be the reason behind this ?

>

> Thanks,

> Mayank

Hi All,

I would like to inform you of a suggested improvement on how Zephyr is used in an application and how to source boilerplate.cmake code.

https://github.com/zephyrproject-rtos/zephyr/pull/23054

Today, each application contains:
include($ENV{ZEPHYR_BASE}/cmake/app/boilerplate.cmake NO_POLICY_SCOPE)

My proposal is to create a Zephyr CMake config package, which allows a user to simply write:

find_package(Zephyr)

in the future.

But let me describe some reason for suggesting this change.

The existing solution works, and for Linux users, it is very common to use environment variables.

However, this solution also have some drawbacks.

One drawback, is that developers working both up-/ and downstream may not be able to always has a single ZEPHYR_BASE, and thus they must still remember to source zephyr-env.sh or run zephyr-env.cmd before working.

Second drawback is people using IDEs such as Eclipse or SES, must ensure that ZEPHYR_BASE is correctly set BEFORE launching the IDE.

If only having one Zephyr, that is not a big issue, but having out-of-tree applications depending on different Zephyr versions or working both up- / downstream does make this a bit more annoying.

Especially many Windows users are not fond of running a command script prior to launching their IDE.

Third drawback, is the fact that ZEPHYR_BASE is not sticky after first CMake invocation, as example, if you try the following:

$ source /tmp/zephyr-1/zephyr/zephyr-env.sh

$ cd <path-to-sample>/hello_world

$ mkdir build; cd build

$ cmake -GNinja -DBOARD=nrf52840_pca10056 ..

-- Zephyr version: 2.2.0-rc2

-- Found PythonInterp: /usr/bin/python3.6 (found suitable version "3.6.8", minimum required is "3.6")

…

Loaded configuration '/tmp/zephyr-1/zephyr/boards/arm/nrf52840_pca10056/nrf52840_pca10056_defconfig'

…

-- Build files have been written to: /tmp/zephyr-1/zephyr/samples/hello_world/build

$ ninja

…

[122/122] Linking C executable zephyr/zephyr.elf

# This could be similar to a user changing folder to do something else,

# or working in an IDE where the ZEPHYR_BASE is different from the value on first CMake invocation.

# and therefore has sourced a different Zephyr

$ source /tmp/zephyr-2/zephyr/zephyr-env.sh

# Indicate a CMakeLists.txt change causing CMake re-run on next ninja invocation.

$ touch ../CMakeLists.txt

$ ninja

[0/1] Re-running CMake...

-- Zephyr version: 2.1.99

-- Selected BOARD nrf52840_pca10056

-- Found west: /home/<user>/.local/bin/west (found suitable version "0.7.0", minimum required is "0.6.0")

-- Loading /tmp/zephyr-1/zephyr/boards/arm/nrf52840_pca10056/nrf52840_pca10056.dts as base

Error: nrf52840_pca10056.dts.pre.tmp:394.23-24 syntax error

FATAL ERROR: Unable to parse input tree

CMake Error at /tmp/zephyr-2/zephyr/cmake/dts.cmake:188 (message):

  command failed with return code: 1

Call Stack (most recent call first):

  /tmp/zephyr-2/zephyr/cmake/app/boilerplate.cmake:460 (include)

  CMakeLists.txt:5 (include)

-- Configuring incomplete, errors occurred!

See also "/tmp/zephyr-1/zephyr/samples/hello_world/build/CMakeFiles/CMakeOutput.log".

See also "/tmp/zephyr-1/zephyr/samples/hello_world/build/CMakeFiles/CMakeError.log".

FAILED: build.ninja

/opt/cmake-3.14.4-Linux-x86_64/bin/cmake -S/tmp/zephyr-1/zephyr/samples/hello_world -B/tmp/zephyr-1/zephyr/samples/hello_world/build

ninja: error: rebuilding 'build.ninja': subcommand failed

$

As seen above, the Zephyr in used is now in mixed state and thus failing.

I have therefore made the following PR, which solve both above mentioned problems, as well as give a better user experience.

https://github.com/zephyrproject-rtos/zephyr/pull/23054

It has the following advantages:

1. CMake will use `find_package` to locate the right Zephyr
2. Version can be used in an application to ensure the correct Zephyr version for Zephyr users when having multiple Out-of-tree apps
3. ZEPHYR_BASE is sticky after first CMake invocation, thus if Zephyr based is changed after first run, it will not impact subsequent CMake invocations (directly or through ninja).
4. Allows to have multiple projects with different Zephyr bases in same IDE, as example Eclipse.
5. Easier getting started for developers. No boilerplate sourcing, no ZEPHYR_BASE / zephyr-env.sh sourcing, just put:
6. Fully backward compatible with existing ZEPHYR_BASE.

So a Zephyr sample such as Hello World may look as:

https://github.com/zephyrproject-rtos/zephyr/blob/a886b9ae428c066bad73cc509757571b70a52d1f/samples/hello_world/CMakeLists.txt
# SPDX-License-Identifier: Apache-2.0

cmake_minimum_required(VERSION 3.13.1)

find_package(Zephyr 2.2.0 HINTS $ENV{ZEPHYR_BASE})

project(hello_world)

target_sources(app PRIVATE src/main.c)

And a downstream user creating an app based upon Zephyr without version check, can simply do:

`find_package(Zephyr)`

If a user wants to support the possibility of overriding the Zephyr by manually specifying ZEPHYR_BASE, the following can be used:  

`find_package(Zephyr HINTS $ENV{ZEPHYR_BASE})`

A user requiring Zephyr version 2.2.0 or above can do:

`find_package(Zephyr 2.2.0)`

A user requiring an exact Zephyr version 2.2.0 can do:

`find_package(Zephyr 2.2.0 EXACT)`

To show that Zephyr base is now sticky, I have re-tried above failure using the new principle:

$ source /tmp/zephyr-1/zephyr/zephyr-env.sh

$ mkdir build; cd build

$ cmake -GNinja -DBOARD=nrf52840_pca10056 ..

Including boilerplate (zephyr base): /tmp/zephyr-1/zephyr/cmake/app/boilerplate.cmake

-- Zephyr version: 2.2.0-rc2

-- Found PythonInterp: /usr/bin/python3.6 (found suitable version "3.6.8", minimum required is "3.6")

…

Loaded configuration '/tmp/zephyr-1/zephyr/boards/arm/nrf52840_pca10056/nrf52840_pca10056_defconfig'

…

-- Build files have been written to: /tmp/zephyr-1/zephyr/samples/hello_world/build

$ ninja

…

[119/124] Linking C executable zephyr/zephyr_prebuilt.elf

Memory region         Used Size  Region Size  %age Used

           FLASH:       11484 B         1 MB      1.10%

            SRAM:        5180 B       256 KB      1.98%

        IDT_LIST:          56 B         2 KB      2.73%

[124/124] Linking C executable zephyr/zephyr.elf

$ source /tmp/zephyr-2/zephyr/zephyr-env.sh

$ touch ../CMakeLists.txt

$ ninja

[0/1] Re-running CMake...

Including boilerplate (zephyr base (cached)): /tmp/zephyr-1/zephyr/cmake/app/boilerplate.cmake

-- Zephyr version: 2.2.0-rc2

…

-- Loading /tmp/zephyr-1/zephyr/boards/arm/nrf52840_pca10056/nrf52840_pca10056.dts as base

…

Loaded configuration '/tmp/zephyr-1/zephyr/samples/hello_world/build/zephyr/.config'

No change to configuration in '/tmp/zephyr-1/zephyr/samples/hello_world/build/zephyr/.config'

No change to Kconfig header in '/tmp/zephyr-1/zephyr/samples/hello_world/build/zephyr/include/generated/autoconf.h'

…

-- Build files have been written to: /tmp/zephyr-1/zephyr/samples/hello_world/build

[86/90] Linking C executable zephyr/zephyr_prebuilt.elf

…

[90/90] Linking C executable zephyr/zephyr.elf

$ echo $ZEPHYR_BASE

/tmp/zephyr-2/zephyr

Also a small screenshot from Eclipse to show that two projects can be imported with different Zephyr bases, and both be built and using their own Zephyr base (and in  this example own requested Zephyr version):

If anyone is interested in a presentation of the new possibilities, I will be happy to make a small presentation for next TSC meeting.

Torsten Tejlmand Rasmussen

Senior R&D Engineer

P: +47 72 89 92 47

Nordic Semiconductor

Otto Nielsens veg 12, 7052 Trondheim, Norway

www.nordicsemi.com

Hi Narendar,

Please create a github bug report, with details on how to build your application and reproduce the issue.

You may share a branch in the GH issue if that is faster for me or others  to reproduce and debug.

Regards,

Vinayak