manjaro linux 18.1 – not ready for virtualization prime time

I read of a new group selling Linux computers, Tuxedo Computers (https://www.tuxedocomputers.com), which is bundling Manjaro Linux with the hardware. All the commentary about Manjaro is laudatory, especially when running on the Tuxedo hardware. So I navigated over to the Manjaro Linux site (https://manjaro.org) and downloaded the latest ISO release, 8.1.

I run a number of Linux systems virtually using Oracle’s VirtualBox on my Mac, all successfully. Before installing a Manjaro guest VM, I updated to the latest VirtualBox, 6.1.2. The update was no problem with my existing VMs, which are a mix of RHEL/CentOS/Oracle Linux 8.1 releases, and various Debian/Ubuntu VMs as well. I was able to update the guest extensions on all of them with no issues whatsoever. Keep that in mind.

Installation of Manjaro was absolutely smooth with no issues. It was attempting to install the guest extensions in the VM where I ran into issues. I tried the regular way of installing them by mounting the extensions CD image and running the installation script. I installed the minimal image, which meant I had to install the kernel header files matching my kernel, gcc, and make. I then ran the installation script and had no failures. On reboot unfortunately the extension that enabled folder sharing of the host’s filesystem failed. The dmesg error indicated a lack of kernel symbols. I then went looking in the Arch Linux forums (Manjaro is based on Arch) and discovered there was an Arch package with those extensions pre-built. Fine. I uninstalled the extensions I installed, then installed using pacman. On reboot the extension still failed to load because the module “taints the kernel”. The message is part of the screen capture at the top of the post.

I won’t put up with this. I would love to run this distribution because its tools are up to the very latest (gcc and Python in particular), and the kernel is the near-latest version 5.4. There’s an awful lot to like about this distribution. But this problem with the VirtualBox kernel modules, especially ones provided via pacman, is a show stopper for me. I need the ability to share files between my VMs and my Mac, or between VMs using the Mac shared folder. That feature works like a charm with every distribution I have except for Arch in general, and this version of Manjaro.

There’s also one other aggravation with Arch/Manjaro I really don’t care for, and that’s the attitude that crops up in the forums, and is exemplified by the comment “that’s not the way we do things in Arch.” Fine. It wouldn’t be such a big deal if it all worked without drama, but it doesn’t. And I’ve already had my run-ins with Arch on my Raspberry Pi (3 and 4) systems as well as an attempt in times past to just get an Arch VM running. The Raspberry Pi installations eventually all failed to properly update after a time, and the VM never worked, including the VirtualBox shared filesystem module.

In April Ubuntu 20.04 will be released with the latest kernel and tools, and I’ll step up to that and call it a day. If I need to keep on the bleeding edge, there’s always the non-LTS Ubuntu releases as well, and there are also ways to keep a Debian installation on leading release tools. I don’t need Manjaro and the Arch attitude that comes with being derived from Arch. All the other distros Just Work. In the future when my clients ask what Linux distribution to install and run, Manjaro won’t be one that I recommend.

I might be retired but I still do a bit of consulting.

notes on using the arduino ide with the adafruit feather nrf52 bluefruit le

Adafruit Feather nRF52 Bluefruit LE – nRF52832

All of this work is done on my MacBook Pro with macOS Mojave 10.14.6 (I haven’t upgraded beyond Mojave yet because I don’t see the need on the MacBook, and if I need macOS 10.15 it’s installed on my Mac Mini).

First, let’s sort out my (miss) adventures with Arduino IDE. It started a few days ago when I decided to do something with the Feather, a device I’d purchased nearly a year prior. I’d plugged it in several times when it first arrived and I’d used the Arduion IDE’s Serial Monitor (Tools > Serial Monitor) to check it. It shipped with the StandardFirmataBLE sketch compiled and installed, so I was able to see the device’s internal configuration on the Serial Monitor as well as test it with Adafruit’s Bluefruit LE Connect app. I then put it aside and wound up forgetting about it until just recently. Adafruit still sells the device today, which is good.

It was during this restart I noticed that my version of the Arduino IDE was at 1.8.7, but the website had released 1.8.10. So I downloaded the latest IDE software and installed it.

After the IDE updated I updated the Feather’s bootloader software. That went off without a hitch. Then I started to look at the sample sketches for the board that Adafruit provides. Since it already came with the Firmata sketch installed and running, I found the source in the samples and attempted to recompile and upload. That’s when everything went screwy. Here’s an example of the error output.

Arduino: 1.8.10 (Mac OS X), Board: "Adafruit Bluefruit Feather nRF52832, 0.2.13 SoftDevice s132 6.1.1, Level 0 (Release)"

/Users/wbeebe/Git/arduino/libraries/Servo/src/nrf52/Servo.cpp: In member function 'void Servo::writeMicroseconds(int)':
/Users/wbeebe/Git/arduino/libraries/Servo/src/nrf52/Servo.cpp:92:12: error: 'g_APinDescription' was not declared in this scope
  instance=(g_APinDescription[pin].ulPWMChannel & 0xF0)/16;
            ^~~~~~~~~~~~~~~~~
/Users/wbeebe/Git/arduino/libraries/Servo/src/nrf52/Servo.cpp: In member function 'int Servo::readMicroseconds()':
/Users/wbeebe/Git/arduino/libraries/Servo/src/nrf52/Servo.cpp:123:12: error: 'g_APinDescription' was not declared in this scope
  instance=(g_APinDescription[pin].ulPWMChannel & 0xF0)/16;
            ^~~~~~~~~~~~~~~~~
Multiple libraries were found for "Firmata.h"
 Used: /private/var/folders/d9/n152_yx519121c0nlmlds71c0000gn/T/AppTranslocation/2101D1AD-76B6-4A4E-A36D-936C2672D030/d/Arduino.app/Contents/Java/libraries/Firmata
Multiple libraries were found for "Adafruit_LittleFS.h"
 Used: /Users/wbeebe/Library/Arduino15/packages/adafruit/hardware/nrf52/0.14.6/libraries/Adafruit_LittleFS
Multiple libraries were found for "InternalFileSystem.h"
 Used: /Users/wbeebe/Library/Arduino15/packages/adafruit/hardware/nrf52/0.14.6/libraries/InternalFileSytem
Multiple libraries were found for "bluefruit.h"
 Used: /Users/wbeebe/Library/Arduino15/packages/adafruit/hardware/nrf52/0.14.6/libraries/Bluefruit52Lib
Multiple libraries were found for "Servo.h"
 Used: /Users/wbeebe/Git/arduino/libraries/Servo
 Not used: /private/var/folders/d9/n152_yx519121c0nlmlds71c0000gn/T/AppTranslocation/2101D1AD-76B6-4A4E-A36D-936C2672D030/d/Arduino.app/Contents/Java/libraries/Servo
 Not used: /Users/wbeebe/Library/Arduino15/packages/adafruit/hardware/nrf52/0.14.6/libraries/Servo
Multiple libraries were found for "Wire.h"
 Used: /Users/wbeebe/Library/Arduino15/packages/adafruit/hardware/nrf52/0.14.6/libraries/Wire
exit status 1
Error compiling for board Adafruit Bluefruit Feather nRF52832.

Lots of error messages. The key to a solution is line 20, but I’ll get to that. In the meantime I went slumming on the Internet for solutions, none of which really worked, so that don’t bear repeating here. What finally got me out of this mess and allowed the sketch to compile is the removal of the libraries folder pointed to in line 20 (which none of the on-line experts never mentioned). I don’t know what happened except that something got out of sync in the upgrade to Arduino IDE 1.8.10, and that caused a cascade effect in the error messages. When I removed the libraries folder and then reloaded the Adafruit nRF52 specific libraries again, everything was just fine. The libraries folder was recreated by the IDE.

In the mean time I decided to try to run a sketch that simply flashed the LEDs on the Feather board. The listing follows.

#include <Arduino.h>

/**
 * Sketch demonstrates mutli-tasking using the Scheduler.
 * Demo creates a loop2() that runs in 'parallel' with loop().
 * - loop() toggles LED_RED every 1 second
 * - loop2() toggles LED_BLUE every 300 milliseconds
 */

void setup() {
    // whb Nov 2019
    // This must be performed first or else nothing else will work.
    // No, I don't know exactly why.
    //
    Serial.begin(115200);

    // LED_RED & LED_BLUE pins are already initialized outputs.
    // Create loop2() using Scheduler to run in 'parallel' with loop()
    //
    Scheduler.startLoop(loop2);
}

/**
 * Toggle the Red LED every second
 */
void loop() {
    digitalToggle(LED_RED); // Toggle LED 
    delay(1000);            // wait for a second
}

/**
 * Toggle the Blue LED every 300 milliseconds
 */
void loop2() {
    digitalToggle(LED_BLUE); // Toggle LED 
    delay(300);              // wait for a half second  
}

There’s far more comments than code, and that’s after not including the boilerplate block comment at the start of the code. This is from an example sketch provided when you install the Adafruit libraries. So I compiled and uploaded it, and the lights didn’t flash. Ok. I went online to Adafruit’s Feather documentation and copied an even simpler sketch that just flashed the red LED. Again, compiled and uploaded and nothing happened. The red LED just glimmered, like it did while the upload to the Feather was taking place.

I went back and looked at the major Firmata sketch, and at setup specifically, and that’s when the light went off. So I added line 15 above, and bingo, the lights started to flash like they’re supposed to. I played with a number of other sketches like this one. Load them without the Serial.begin() and nothing worked. Add that single line and it worked as advertised.

I also like this sketch because it points to the use of freeRTOS in the code. Keep in mind that the processor in the feather is a 32-bit ARM Cortex M4 processor running at 64MHz with 64K of RAM and 256K of Flash. It’s more than enough for basic embedded work, including freeRTOS. When I think back to the late 1970s and early 1980s I would have given anything to have this back when all that was available were Commodore 64s and Apple ][s and TRS-80s and the original IBM PC. This even blows the original Macintosh away with its 8MHz Motorola 68000. And here it all is now, on a board not much bigger than a stick of chewing gum and all for just $25 instead of the hundreds to thousands back then. Anyway…

The Arduino and its environment are now stable again, and I can build (in C, and possible C++) “sketches” for the Feather. This goes along with my Circuit Playground and my little Adafruit 32U4 powered robot. All of my code is checked into my Github location, https://github.com/wbeebe/arduino