Brave testers with eboards needed for V4 testing

[Johan Sjöblom]

We would at some point like to upgrade the Bluetooth functionality in picochess from the existing Bluepy to the new Bleak.

To do that I took the first step and made preparations in all the eboard software code. In theory these preparations should not cause any problems. But we do need to test all the chess boards so that we know we have a stable base before porting to Bleak.

I will test on my DGT board and on my DGT 3000 clock.

At the moment you need the skill to checkout a branch to be able to test.

The PR: https://github.com/JohanSjoblom/picochess/pull/171

The branch to test: 170-eboard-to-async

You can report any strangeness here, or add a comment on this issue: https://github.com/JohanSjoblom/picochess/issues/170

-- Johan

[waldomiro dias machado junior]

Dear friends,

I am a regular user. I am able to test the Certabo Bluetooth Classic and the Chessnut Air. 

If you wish, I am available. 

I have a Raspberry Pi 4 and a Pi 5 connected to two 10.1-inch monitors. 

Waldomiro

Em sábado, 29 de novembro de 2025, Johan Sjöblom <messi...@gmail.com> escreveu:

Em algum momento, gostaríamos de atualizar a funcionalidade Bluetooth do picochess, substituindo o Bluepy atual pelo novo Bleak.

Para isso, dei o primeiro passo e fiz preparativos em todo o código do software do tabuleiro eletrônico. Em teoria, esses preparativos não devem causar problemas. Mas precisamos testar todos os tabuleiros de xadrez para termos certeza de que temos uma base estável antes de migrar para o Bleak.

Vou testar na minha placa DGT e no meu clock DGT 3000.

No momento, você precisa da habilidade de acessar uma branch para poder realizar testes.

O PR:  https://github.com/ JohanSjoblom/picochess/pull/ 171

O branch a ser testado: 170-eboard-to-async

Você pode relatar qualquer comportamento estranho aqui ou adicionar um comentário nesta questão  : https://github.com/JohanSjoblom/picochess/issues/170

-- John

Você
recebeu esta mensagem porque está inscrito no grupo "PicoChess" do Google Groups. Para cancelar sua inscrição neste grupo e parar de receber e -
mails dele, envie um e-mail para picochess+...@googlegroups.com .
Para visualizar esta discussão , acesse https://groups.google.com/d/msgid/picochess/6af00dc6-8a7a-4596-bd12-6184aad5951cn%40googlegroups.com .

[Aldo Bleeker]

I'll be checking out the Millennium eOne, and the Millennium The King Competition using the ChessLink (the ChessLink of the eOne is built-in).

Aldo

[Johan Sjöblom]

Instructions for the brave, courtesy chatgpt, by using a terminal:

To go to this branch and test:

cd /opt/picochess
git fetch origin
git checkout 170-eboard-to-async
git pull --ff-only origin 170-eboard-to-async
sudo systemctl restart picochess

To get back to master (latest published ok version):

cd /opt/picochess
git checkout master
git pull --ff-only origin master
sudo systemctl restart picochess

[Luigi Vogliobene]

Hello,

I am able to test the Chessnut Pro bluetooth.

Luigi

[Aldo Bleeker]

As I'm not an expert on Git, does this pull PR #171?

Aldo

[Johan Sjöblom]

Yes, the PR is more like a ticket. The content is the branch, so yes

It will pull 170-eboard-to-async

Which is the PR 171

[Aldo Bleeker]

I was watching the PicoChess RPi 5 start, and it looked fine. But then I noticed that I hadn't even switched on the Millennium eOne. Since it wasn't even turned on, I'm pretty sure that there was at that moment no bluetooth connection, but the usual message 'no ChessLink e-Board' message accompanied by the spinner didn't show, and after 10 minutes or so still hasn't.

I switched the board on, and then the message popped up! It disappeared, and the connection was working.

I broke the connection by switching off bluetooth and the 'Connection to Bluetooth peripheral lost: [Errno 32] Broken pipe' message was displayed after a while. I switched bluetooth on again, but the connection wasn't restored, and after a couple of minutes still isn't. The message hasn't gone away, and PicoChess isn't responding to the movement of pieces.

It hasn't frozen, going into the menu is working, and leaving the menu removed the message. It still isn't responding to moves though.

Switching bluetooth off, and on again hasn't restored the connection.

Restarting the picochess-service did work, and it started to move.

I then started PicoChess again, and this time it connected immediately, and PicoChess responded, moves were registering and displayed.

Breaking the connection for just a showrt while, less than 10 seconds, wasn't an issue, the connection was restored and I could resume the game.

Aldo

[Johan Sjöblom]

Thanks for testing. Yes sometimes the connection break cannot recover and sometimes it can recover. And sometimes the web socket also breaks and then you do not get the error messages before the connection restores. If you send me the /opt/picochess/picochess.log file I can analyse in more detail.

When you start picochess you should not hear it say ”OK” before there is a connection. Does that work on your board?  Example, when you forgot to turn the board on, it should not have said OK, just ”Engine setup”.

[Luigi Vogliobene]

Hello,

tested w/ Rpi5 & Chessnut Pro eboard : all is OK, very fast coupling bluetooth.

Luigi

[Francois Vannier]

Hi Joahn & all

Tests on rpi4B with DGT RevII:

• Connexions are OK, most of the time very fast, sometimes a couple of seconds, but no failure
• I got this following sequence, maybe not related to this very branch, debug log attached:
• Cold start both eboard & Picochess until "ok"
• Switch off the eboard
• Switch back on the eboard. The display on the revII is showing an empty board (all pieces on the board were expected) & time = 00:00, as if not connected. The keys of the local web I/F are inoperant from now on
• I can navigate thru the engine menu using the RevII keys. The current menu entry is displayed on both RevII & local web
• Selecting & starting an engine is possible until "ok" message, with still an empty board on the RevII display & no key working on the local web I/F
• Piece movement is not detected. Navigating thru the engine menus is still possible, but it's impossible to have access to other menus like 'system'
• If I try to navigate into the local web menu when the board is off, the first key pressed is registered & the menu changed, but impossible to go further. When the eboard is back on, both menu displays are no longer synchronized, like one showing 'Engines' & the other "System'

Francois

--
You received this message because you are subscribed to the Google Groups "PicoChess" group.
To unsubscribe from this group and stop receiving emails from it, send an email to picochess+...@googlegroups.com.
To view this discussion visit https://groups.google.com/d/msgid/picochess/b4407060-a1df-4149-bfd2-67c54a943e4en%40googlegroups.com.

[Aldo Bleeker]

RPi5, bluetooth ChessLink, Millennium eOne.

I started PicoChess with bluetooth switched off. PicoChess started, but the message 'no ChessLink e-board' and spinner didn't show. Christina says 'Picochess Engine Setup', but she doesn't say 'OK'.

Switched bluetooth on, and waited a couple of seconds. Now Christina says 'OK'. I played some moves, and PicoChess is responsive, so there's a bluetooth connection.

I switched bluetooth off, and waited 5 seconds or so. PicoChess displays a message, 'Connection to Bluetooth peripheral lost: [Errno 32] Broken pipe'. I switched bluetooth on again. The message stays in place, PicoChess isn't responsive, the bluetooth connection hasn't been restored.

I restarted picochess.service, PicoChess connects, a new game starts and PicoChess is responsive.

Picochess connects quickly every time if the board is already switched on, and even when it's switched on later.

It doesn't say 'OK' when the board doesn't connect, but the familiar 'no e-Board' message to indicate that is missing. The RPi5 doesn't have a speaker connection, that's why I usually use it without sound, so the message is quite useful.

A lost connection isn't restored every time, but restarting the service does work, even though it restarts the game.

Aldo

[Johan Sjöblom]

Ok brave testers. Thank you Luigi, Aldo, and Francois.

On the DGT boards (like RevII) there has been a long time problem with logs filling up with "(ser) clock is locked over 2secs". I have seen it myself many times, now thanks to your log Luigi I think I know what happens. Whenever we send a clock command to the DGT picochess code sets "a lock". It clears the log only if it gets an ACK back from the board. If that goes missing, then its stuck. I have now added a check if we are missing the ACK for the clock commands on DGT. That _could_ and _should_ help.

On the Chesslink board I have now changed also the "no e-board" message into asynchronous which should make it appear. I also consider this message important. On the DGT board I had already made it async.

if you want to retest on DGT RevII and Chesslink please update your code by doing a pull like this:

> cd /opt/picochess

> git pull

It should say that you got a new board.py version. If you then restart the picochess like this you can test again when you have time.

> sudo systemctl restart picochess

You can also do an update from the menu if you dont have the possibility to go in via ssh or have a keyboard attached.

PS. There is still a problem I am looking for, and that is the strange thing that when we break the Bluetooth connection by turning of the boards the web socket seems to get problems and sometimes the Web interface freezes. As soon as Antonio is ready with the next version of the web client I think I will go in there and add some keep-alive to the web connection. DS.

[Johan Sjöblom]

After some serious Raspberry troubleshooting I found out the following: If you are using Chromium and Wifi on 5GHz AND you abruptly turn off a Bluetooth device all chromium sessions stop working, INCLUDING chromium browsers on other machines that point to the Raspberry Pi ip address. I checked what happens and the local Pi Web server is perfectly fine, but you cannot use any Chromium browser to connect to it anymore. So now I am going to test with 2.4GHz wifi connection and Firefox to see if that works any better...

[Aldo Bleeker]

I pulled the latest PR, and I got a new board.py. Wit this version I get the message when I break the connection. I'm not sure if it's a coincidence, but restoring the connection seemed to work a lot better as well.

I don't get the message when I start PicoChess with bluetooth turned off, right?

Aldo

[Johan Sjöblom]

Aldo, I thought you should have got the "no e-board" message also when you turn on the Pi with the board in power-off... 

And yes, restoring the connection now has retries...

[Johan Sjöblom]

And to all some important information on if you have Raspberry Pi3 or Pi4, or 400. DO NOT USE Wifi 5Ghz. You can look up the following if you find more information:

"The BCM43455 Wi-Fi/Bluetooth coexistence logic behaves differently between 2.4 GHz and 5 GHz"

Co-existence is a problem if you abruptly turn off a Bluetooth device connected to Pi3, 4. The problem goes away if use a 2.4 GHz Wifi connection (with a fresh reboot).

The problem is this chip: Broadcom BCM43455 Wi-Fi/BT combo chip (also called CYW43455). You can also use a bluetooth dongle, but the onboard will get problems when BT gets problems.

[Aldo Bleeker]

That's what I thought too.

That's why I restarted the RPi, and checked if I got the PR (I did). I've tried again, but I didn't get the message. I turned on bluetooth, I saw the message, and PicoChess connected.

Aldo

[Johan Sjöblom]

In plain English: If you don't see the "no e-board" messages when you have turned off the power off your eboard or lose connection because of something else: Use 2.4GHz Wifi on your Pi, not 5GHz.

On my Pi4 it takes about 10 seconds before the operating system notices that the DGT board has been switched off. Once Pi OS detects the disconnect (I get a popup screen) --> the spinner comes up on all browsers I have pointing to the Pi4. I can then turn the DGT board back on and all browsers wake up in about 2-5 seconds and I can continue playing on my board.

[Johan Sjöblom]

Aldo, I made another fix that should give you the "no e-board" message on a cold start with Chesslink turned off from start.

To get this fix, you need to do another git pull, and a picochess restart. Lets see if this small fix would do it.

[Aldo Bleeker]

Mines an RPi 5, not a 3, 4, or 400. Anyway, I've applied the PR, and switched off wifi altogether. Still no message though.

Aldo

[Aldo Bleeker]

There's no message when I start PicoChess with bluetooth off. When I switch on bluetooth, nothing seems to happen for about 10 seconds. Then the message is displayed (spinner is spinning). This lasts about another 10 seconds, it disappears, and the connection is working.

Aldo

[Johan Sjöblom]

Ah, clarification. Do you have Bluetooth on on Pi, and you start with the board in power-off, And you do not get any messages?

[Aldo Bleeker]

Bluetooth is on on the Pi, but I start PicoChess with the board (and bluetooth) switched off, so at that time PicoChess sees there's 'no ChessLink e-Board', which is quite right. This is the message I see at the place of the clock when I start the current master branch, until a bluetooth connections is established, then the message is replaced by the name of the engine, and the clock. But when I start the 170 branch, I immediately get the engine/clock, there's no indication that the bluetooth connection hasn't been established yet. Then when I switch on the board (and bluetooth), the connection is established, but the engine/clock doesn't change. This way you can't tell that PicoChess has started without bluetooth connection, but now has established one. This new, earlier you could quite clearly tell by the 'no e-Board' message that is displayed right away when you start PicoChess.

Aldo

[Aldo Bleeker]

The issue may not be clear, so I'd like to add that starting PicoChess from the 170 branch with the board turned on seems to establish a bluetooth connection straight away. It looks like it's connecting quickly and reliably, the PicoChess responds to moves made on the board, and is available when you start a game. And when the connection is lost, it now is able to restore it. This is all rather new, but it's already working well.

The only issue in this branch is that when the board is switched off when PicoChess is started and it can't establish the bluetooth connection with the board, no message is displayed, and in the master branch you get the 'no e-Board' message. And in that case a message stating the connection can't be established is quite useful.

Aldo

[Johan Sjöblom]

Thank you Aldo for the clarification. I will need to make another attempt to fix the startup "no ChessLink e-Board" if ChessLink connection cannot be established at startup. I thought the previous change should have done it, but obviously I have to try one more time :-) I will be back!

[Johan Sjöblom]

OK, ChessLink cold start fix number 2: If you do a git pull you should now get a new eboard/chesslink/board.py update.

[Johan Sjöblom]

Do not be surprised if you pull and get a lot of new web client files. Both this branch and master branch now contains Antonio's latest web client resolution improvements.

[Aldo Bleeker]

Hi Johan,

I'm using the latest PR's, and I'm now seeing the 'no ChessLink e-board' when I start PicoChess without bluetooth connection. When the connection is lost during a game, and is back again 10 seconds or so later, the connection is restored, and I can just continue the game. Great job!

By the way, PicoChess is saying PicoChess Version 4.1.7, because it says so in clock.html.

Aldo

[Johan Sjöblom]

Nice. We managed to get ChessLink to work at least as good as master, with potential for future improvement when we start using Bleak library.

I will fix the version...  It must have come from Antonio's code... Good catch!

[RandyR]

Is this correct? The way I read it is if you use the 2.4 GHz band for Wi-fi it could cause interference with BT, which only operates at 2.4 GHz. So the recommendation is to use 5 GHz Wi-fi to prevent this (or at least minimize it). 

[Johan Sjöblom]

You should not use the 5Ghz Wifi and Bluetooth. Its not because of radio interference, it is because of the broadcomm chip.

The chip on many Pi3 and Pi4 can stop chromium browser from working with the web server when you have Bluetooth devices that disconnect abruptly.

I recommend to use the 2Ghz wifi with Bluetooth!

--

You received this message because you are subscribed to the Google Groups "PicoChess" group.
To unsubscribe from this group and stop receiving emails from it, send an email to picochess+...@googlegroups.com.

To view this discussion visit https://groups.google.com/d/msgid/picochess/9f68d9ce-1166-4c4f-bc88-07ae7e14ff24n%40googlegroups.com.

[Randy Reade]

Are you talking about the local kiosk browser or a remote browser? And only Chromium? I haven't noticed any issue on the Pi4 but I don't typically disconnect the board while using it.

To view this discussion visit https://groups.google.com/d/msgid/picochess/CAEk%2BOSS0CXLtj1y%3D-Co4Q7TZd1P%2BxE4iWiehiV%3D563bU0K7i9g%40mail.gmail.com.

[Johan Sjöblom]

I only tested Chromium browser and curl "pings", and Chromium froze on both local Pi4 and on my remote Debian laptop.  That was indeed a big surprise. The Chromium on the remote laptop did not recover either before I had rebooted the Pi4.

The test case I used was to turn of the power of the DGT board in the middle of a game. Note that I am testing in the branch 170-eboard-to-async. The idea was to make sure that we get a spinner showing as soon as we lose connection, not only if the connection is missing during boot, but also if the DGT board connection is lost in midgame. I noticed that the web UI froze when I turned off the power. First I thought that I still had some server code issue with the picochess web server. But then I proceeded and tested a little bit more extensively. I set up a curl session to ping the web server both on the local Pi4 and from the Debian laptop. I also opened a Chromium session on both local and remote. Not always, but in 80% of the cases the abrupt turning off the DGT board in midgame made both curl sessions and both Chromium sessions freeze (both local and remote).

Now when I have the freeze situation I opened 2 new curl sessions in new windows, one local and one remote. These two new sessions both worked proving that the Python Flask web server was still up and running and responding normally. The two original sessions were still frozen. But why had both my Chromium frozen, and even worse, restarting Chromium after making sure I had killed all Chromium processes still showed a frozen web UI in Chromium: Both on local and remote. What a surprise. But as said, the new curl sessions were ok. Only a Pi4 reboot could clear the bad state in the broadcom chip caused by a combination of having 5GHz Wifi and an Bluetooth connection that disconnects abruptly.

There is a difference in the handling of Wifi 5Ghz and 2.4GHz. I then switched my Pi4 to 2.4GHz Wifi and redid the test --> No frozen Chromiums any more, neither locally nor remotely. And the spinner came up as soon as the Pi4 had detected loss of the Bluetooth connection (which on my Pi takes about 10 seconds). As soon as the Pi OS has detected loss of Bluetooth the spinner started showing on both local and remote Chromium browsers. Just because I switched my Pi4 to 2.4GHz Wifi instead of my 5GHz.

I have finished testing this branch so I will merge it to master. Its the first step towards upgrading bluepy to Bleak... and maybe with Bleak all board handling would be faster as Bleak supports async.

[Randy Reade]

Thanks for the details. I want to see if I can duplicate it. I can test both bluepy and classic Bluetooth (DGT). A quick initial test using Chesslink on v3.4 Trixie with a Millennium board, when turning OFF the Chesslink device in the middle of a game and losing the connection, turning it back ON the connection never recovered (the Pi never reconnected to the board). I didn't check if the web interface was frozen but I'm using 5GHz wifi to test on a Pi4. I will switch to a v4 card and continue testing there.

To view this discussion visit https://groups.google.com/d/msgid/picochess/2456ed00-01c6-439c-8c3a-e53131c77b23n%40googlegroups.com.

[Johan Sjöblom]

I  have now merged this, and 2 other branches into master. So to test you can just use master branch now. A git pull should bring you all the new stuff.

[RandyR]

Testing the Chesslink connection I notice that when turning the board OFF (actually the Chesslink device which is powering the board) after PicoChess is running and connected, the BLE connection sometimes takes a while to connect. Perhaps bleak will help here.

I notice the "spinning wheel" doesn't spin and the message just shows "no ChessLink e-Board" + " \" or "-", etc. Not a deal breaker.

Would it be possible to restore the White and Black times to the point where the connection was lost?

I'm testing master branch, Pi4 in 1024x600 display case, using 5GHz WiFi with Chromium kiosk and Firefox on laptop. Menu functionality is working when board is disconnected.

Randy

[Johan Sjöblom]

I was thinking about the player clock times as well... It would be good if we could detect the Bluetooth connection loss immediately... At latest at the time point when we do detect it; I was thinking we should put the clock in "pause" mode... Or as you say, record the times and put them back.... Not sure how easy it would be to stop the clock... or record the times, but that would indeed be nice.

The most annoying thing is that it takes 10-12 seconds (on my Pi4) before the Pi OS has detected a loss of Bluetooth... Not sure if Bleak will help, but if it doesnt we should have a "keepalive" of some kind to detect it on a second level in picochess, not 10 seconds...  Not sure why the Pi takes 10 seconds to detect a lost signal, maybe its just because its so "abrupt" loss when you turn off the power?

[Johan Sjöblom]

Some more details on which hardware on Pi3/Pi4 can give problems on Chromium if a bluetooth device abruptly disconnects. The chip is BCM43455. Check

> lsmod | grep brcm

If you see brcmfmac in the answer, like for example this one, then that hardware can cause cross-over problems from Bluetooth to 5GHz Wifi.

brcmfmac            327680  0
brcmutil               16384     1 brcmfmac
cfg80211             843776   1 brcmfmac

Some potential sources:

https://github.com/raspberrypi/linux/issues/7092
https://github.com/RPi-Distro/firmware-nonfree/issues/8
https://discourse.osmc.tv/t/bluetooth-and-wlan-at-the-same-time/81002
https://raspberrypi.stackexchange.com/questions/104575/raspberry-pi-4-native-wifi-ac-5ghz-not-working
https://arxiv.org/abs/2112.05719

Another VERY important detail, the Pi4 on which I detected the problems with 5GHz I am still running Bookworm, not Trixie. Plenty of fixes has been made in the stack on Trixie.

And below is even more details on my frozen 5GHz when disconnecting the DGT board:

• I am using a Raspberry Pi 4, which contains the Broadcom BCM43455 Wi-Fi + Bluetooth combo chip.

• Wi-Fi is connected on 5 GHz (not 2.4 GHz).

• A Bluetooth LE device (the DGT chess board) is connected.

• The Bluetooth device is disconnected abruptly (power cut) during mid-game.

• Chromium or Firefox was running in the GUI session of Raspberry Pi OS, and on a remote Debian laptop.

• The GUI applications rely on glib/GIO’s NetworkMonitor and BlueZ integration.

• The abrupt Bluetooth disconnect causes a small internal coexistence hiccup in the Wi-Fi/BT chipset or a bad D-Bus network state transition.

• Chromium’s network stack enters a poisoned or broken state, failing to create new TCP connections, even to 127.0.0.1 (and localhost).

• Raw-socket tools like curl and Python sockets continue working, proving that the OS networking layer itself is fine.

-- Johan

[RandyR]

Using 'tail -f /opt/picochess/logs/picochess.log' on DEBUG log I can see that it takes about 6 seconds before bluepy reports a disconnect.

2025-12-06 14:30:45.971 WARNING chess_link_bluepy - worker_thread: Bluetooth read error [Errno 32] Broken pipe

[Johan Sjöblom]

Ah, with this info I could actually design a code change that would detect the connection break earlier....

[waldomiro dias machado junior]

Dear friends, 

Certabo works perfectly on PicoChess V.3.3 and V.4.1.8 via USB cable on my 10.1-inch monitor with a Raspberry Pi 4, but unfortunately it doesn't work on my other monitor with a Raspberry Pi 5 in the same versions 3.3 and 4.1.8; all 64 LEDs on the Certabo blink intermittently. 

Chessnut works perfectly on the Pi 4 and Pi 5! 

Thank you, Waldomiro

Em domingo, 7 de dezembro de 2025, Johan Sjöblom <messi...@gmail.com> escreveu:

Ah, com essa informação eu poderia realmente projetar uma alteração no código que detectaria a quebra de conexão mais cedo...

No sábado, 6 de dezembro de 2025, às 22:35:39 UTC+2, RandyR escreveu:

Usando o comando 'tail -f /opt/picochess/logs/picochess.log ' no log de DEBUG, posso ver que leva cerca de 6 segundos para o bluepy reportar uma desconexão.

2025-12-06 14:30:45.971 AVISO chess_link_bluepy - thread de trabalho: Erro de leitura Bluetooth [Errno 32] Pipe quebrado

No sábado, 6 de dezembro de 2025, às 11:59:48 UTC-6, messi...@gmail.com escreveu:

Eu também estava pensando nos horários do relógio do jogador... Seria bom se pudéssemos detectar a perda de conexão Bluetooth imediatamente... No máximo no momento em que a detectássemos, eu estava pensando que deveríamos colocar o relógio em modo de "pausa"... Ou, como você disse, gravar os horários e depois exibi-los novamente... Não tenho certeza de quão fácil seria parar o relógio ou gravar os horários, mas isso seria realmente ótimo.

O mais irritante é que leva de 10 a 12 segundos (no meu Pi4) para o Pi OS detectar a perda de Bluetooth... Não sei se o Bleak vai ajudar, mas se não ajudar, deveríamos ter algum tipo de "keepalive" para detectar a perda em um segundo nível no picochess, e não depois de 10 segundos... Não sei por que o Pi leva 10 segundos para detectar a perda de sinal, talvez seja porque a perda é muito "abrupta" quando se desliga a energia?

Você
recebeu esta mensagem porque está inscrito no grupo "PicoChess" do Google Groups. Para cancelar sua inscrição neste grupo e parar de receber e -
mails dele, envie um e-mail para picochess+...@googlegroups.com .

Para visualizar esta discussão , acesse https://groups.google.com/d/msgid/picochess/a68fe9b8-8538-4a1f-a52f-359ec2162966n%40googlegroups.com .