[Bug] EV3 firmware installation sends bad status values

[laurensvalk]

During EV3 firmware installation, variations of the following were observed on multiple operating systems from different users. Including Windows and Ubuntu. For some users this does not happen.

Error: EV3 reply status error: 126 for command 246
    at sendCommand (http://localhost:3000/static/js/bundle.js:22082:26)
    at sendCommand.next (<anonymous>)
    at handleFlashEV3 (http://localhost:3000/static/js/bundle.js:22086:42)
    at handleFlashEV3.next (<anonymous>)
    at next (http://localhost:3000/static/js/bundle.js:74363:29)
    at currCb (http://localhost:3000/static/js/bundle.js:74447:7)
    at chCbAtKey (http://localhost:3000/static/js/bundle.js:73961:9)
    at currCb (http://localhost:3000/static/js/bundle.js:74447:7)
    at takeCb (http://localhost:3000/static/js/bundle.js:73793:5)
    at Object.put (http://localhost:3000/static/js/bundle.js:73522:5)

To reproduce, go to https://labs.pybricks.com/ and attempt EV3 installation or recovery. For installation, you'll need to provide your own ZIP file since there isn't one included yet. If you don't have a firmware zip file, just try recovery instead.

[laurensvalk]

Could this be the issue on USB 3.0 systems, worked around in https://github.com/pybricks/pybricksdev/blob/master/pybricksdev/connections/ev3.py#L251-L257 ?

Since indeed:

GET_VERSION = 0xF6 // 246

[laurensvalk]

I am able to flash with these hacks:

diff --git a/src/firmware/sagas.ts b/src/firmware/sagas.ts
index a25e5b26..92efc009 100644
--- a/src/firmware/sagas.ts
+++ b/src/firmware/sagas.ts
@@ -1192,13 +1192,21 @@ function* handleFlashEV3(action: ReturnType<typeof firmwareFlashEV3>): Generator
             ];
         }
 
+        if (command === 0xf6) {
+            // TODO, we know from Pybricksdev that we could still get the version out of this.
+            // But at least don't raise on a bad status.
+            console.log(reply.payload);
+            return [new DataView(new ArrayBuffer(8)), undefined];
+        }
+
         if (reply.status !== 0) {
-            return [
-                undefined,
-                new Error(
-                    `EV3 reply status error: ${reply.status} for command ${command}`,
-                ),
-            ];
+            // TODO: Figure out which status are reliable.
+            // return [
+            //     undefined,
+            //     new Error(
+            //         `EV3 reply status error: ${reply.status} for command ${command}`,
+            //     ),
+            // ];
         }
 
         return [new DataView(reply.payload), undefined];

We don't need to go quite this lenient on errors, though. I think it fails only on certain messages so we don't need to pass them all.

[laurensvalk]

@afarago, are you able to reproduce issues similar to the following? Did you run into this when making your own experimental installer?

If yes, would you be interested to look into this, and possibly implement appropriate workarounds?

[afarago]

I have not seen such problems testing the code through Mac and Win machine.

Any specific way to reproduce it, maybe also on non-Ubuntu systems?

[Sirus-7]

Testing on my Raspberry PI 5 with Debian GNU/Linux 13 (trixie) (64-Bit)
Chromium Version 141.0.7390.122
Visual Studio Code Version: 1.105.1
Python V3.13.5
pybricksdev v2.3.0

EV3 Brick HW: V0.6
Pybricks Beta v2.8.0-beta.2 (https://labs.pybricks.com/)

• Install Pybricks Firmware (local-build-v3.6.1-550-g5b6b4d2a) --> OK
• Connect using USB / disconnect --> OK
• Testing Programs all 5 slots- Run/Stop --> OK
• REPL --> OK
• Restore Firmware V1.06E --> OK

Everything is fine. This is how it can go on.

[laurensvalk]

This was also reported by @pepijndevos:

Error: EV3 reply status error: 127 for command 246