Have a look here: esptool write_flash: error: argument : Address "{upload.erase_cmd}" must be a number

This problem seems to be caused by the esp32 2.0.5 update.

I don't think M5Stack are to blame, this was not a mandatory setting before.

After this update was installed I had the same problem with some of my M5Stack boards, but not all of them. A menu setting for erasing flash before upload or not must now be set. It seems that this menu setting must be sent to esptool, if this is not provided then esptool fails with the Address "{upload.erase_cmd}" must be a number error
There is a simple fix, if you can find your hidden arduino boards.txt file.
The path to the file will be different on Windows, Linux and OSX.
You need to add the following to boards.txt for ALL types of M5Stack/M5Stamp boards and then uploads should work again:

<board_name>.menu.EraseFlash.none=Disabled
<board_name>.menu.EraseFlash.none.upload.erase_cmd=
<board_name>.menu.EraseFlash.all=Enabled
<board_name>.menu.EraseFlash.all.upload.erase_cmd=-e

Replace <board_name> with the correct name for each type of board in the complete list. You can see what the name should be in the section for each M5Stack board.

After editing and saving boards.txt in the correct place, you need to restart the arduino IDE. You will then see in the Tools menu Erase All Flash Before Sketch Upload: Disabled/Enabled. This is the missing required setting for esptool.

Make a copy of your boards.txt before you make any edits so you can recover from any incorrect changes.

Good Luck!

I fixed the problem with the cam2core on the core2. So the UnitCam will work with Arduino IDE code and the core2.

It is a very simple fix:
Replace the first line #include <M5Stack.h>
with #include <M5Core2.h>

Initialising the core code on a Core2 is not the best idea!
Now it works on my Core2 showing the video stream from a UnitCam connected to PortB (pins 26, 36).

Confirmed, the UnitCam does NOT work at all on the Core2 in UIFlow or Arduino.
The Arduino UART to Core does NOT work on the Core2 - it does not boot.
I managed to fix this problem!
I will just use my UnitCam on my Cores and wait until it gets fixed for the UIFlow Core2 - if ever.

I can confirm that the UnitCam does NOT work with UIFlow 1.9.8 on a Core2.
It does not seem to be supported.
A UnitCam stream works without a problem connected to PortB with UIFlow 1.9.8 on an original Core. I have not properly tested the cloud functionality - it didn't seem to work for me. QR code for the url was never displayed.
There must be some technical reason why the UnitCam stream doesn't work on a Core2. Different UART behaviour?
However there is NO warning on the UnitCam store and doc pages that it currently does not work in UIFlow 1.9.8 on a Core2. If you only have a Core2 I don't think it will work. Luckily I have both Core2 and origjnal Cores so I could get it working.
Has anyone successfully tried a UnitCam on a Core2 with Arduino instead of UIFlow?

I cannot run the latest M5Burner with Ubuntu 18.04. It complains that glibc 2.28 is not found.
Unfortunately glibc 2.27 is the most up-to-date version on my system. I cannot upgrade to a newer Ubuntu.
Earlier versions of the M5Burner worked OK. I upgraded to be able to burn firmware on the latest M5 boards like M5Stamp.

This exact problem has been fixed before, see below. Can you please rebuild the M5Burner depending on earlier versions of the system libraries?

Re: Cant run M5Burner in Linux

I've just tried using my atom tailbats with a new OLED Unit and spent a long time wondering why the tailbat grove port didn't seem to work with the OLED.
The USB power and grove connector have really bad electrical connections on my newer matrix atoms (with diffused plastic over the LEDs). This stops the pass-through grove port from working properly unless I press the atom down onto the tailbat quite hard and hold it. The USB connector for the atom power usually works fine, but the atom grove pins don't connect cleanly to the tailbat. If I press too hard then the USB power connection fails and resets the atom (obviously). This is a really bad mechanical design which doesn't use a proper grove socket, it is done with a similar 4-pin connector socket with the correct pin pitch, but it is not grove.

I have just tried a couple of older Atom Matrix cores (with clear plastic over the LEDs) and the grove connector pins on the older Matrix Atoms are very slightly longer and make a better connection. With those, the OLED shows my test animation every time. It also works with an atom lite.

This is not good as I don't know how I can make all the USB power pins and grove pin connections reliable without touching or holding anything. Both USB power and all the grove pins need to work and both of the different connectors need to be in exactly the correct position at the same time. You can't really adjust anything on the tailbat connectors, they are fixed in place. It's difficult to see how the connections could be done with 'proper' grove and USB extension cables to the battery instead of the direct dual connectors. Putting some sort of kludge board between the atom and the tailbat is also a bit of a ridiculous solution, the tailbat and pass-through grove connector should really work on any atom, every time. I could just use the built-in grove port and a USB powerbank (if I can make it stay powered-on all the time).

Anybody else have anything similar? There are some other posts here that tailbats were not working properly, I wonder if some of those problems were actually the same bad mechanical connection issue?

@felmue It was "unconnectable" for about 24 hours, which was extremely annoying. The thing was that the devices could connect to the server and showed green connections. But the UIFlow server couldn't connect back to the devices.
I didn't change anything at my end and now everything is working as before.
I was trying to test the OLED Unit because of a brightness problem and that new unit is not yet supported in the desktop (offline) UIFlow IDE, so I had to be online and connected to try things out.

When I use the OLED unit with UIFlow 1.7.8 on a m5stack atom, the display is quite dim. It looks more like black and grey, not black and white.

It doesn't have good contrast, it looks like it is not getting enough power, but maybe it's actually the display initialisation that is the problem.

When I use arduino and the same m5stack atom with the adafruit sh110x arduino library, the same OLED unit is vey bright and crisp, definitely black and white, with very good contrast.

I suspect the OLED contrast is not getting set correctly during the init.

There are no blocks in UIFlow to adjust the contrast so it cannot be changed.
In arduino I can change the contrast (setContrast) and I can make the display darker or brighter.

Is there something wrong or incomplete with the 1.3" OLED unit display initialisation in UIFlow?

It seems to be working again, my browser now gets a connection to my device. So it was a server-side problem, nothing to do with our m5stack devices, firmware or API keys.