DIV-33 Update firmware via websocket command

[krksgbr]

Changes at a high level:
The messaging protocol was extended to allow receiving a command initiating a firmware update. The command carries a base64 encoded binary of the firmware which is decoded and sent to the Senso. During the update process the driver sends progress messages back to the client.

The discovery process was re-factored to allow simultaneous discovery of Sensos that could be in both "Application" and "Bootloader" mode. The discovery process now also checks the mode txt entry to determine what state a Senso could be in, as older versions of the firmware announce themselves as _sensoControl._tcp instead of _sensoUpdate._udp when in bootloader mode.

Checking the authenticity of the firmware image is still to be added.

Checklist

• Changelog updated
• Code documented

[krksgbr]

The latest changes should make the procedure a lot more reliable. It still doesn't seem bulletproof however. With this stress test I was able to send my controller into a state from which it can only be recovered by power cycling. After a few runs, the controller no longer announces itself via mdns and is also not responsive to commands. It's not typical usage, so I'm not sure how worried we should be about this, but I thought withstanding this would be a good indication of robustness. It's unclear if the conditions that bring this about are specific to the chained updates or if they can arise under normal use as well. In any case, the controller can be recovered by power cycling, so it's a relatively benign problem. We'll have to make sure to communicate this well in the UI.

diff --git a/src/dividat-driver/firmware/main.go b/src/dividat-driver/firmware/main.go
index e0465d8..1ea09b6 100644
--- a/src/dividat-driver/firmware/main.go
+++ b/src/dividat-driver/firmware/main.go
@@ -38,23 +38,30 @@ func Command(flags []string) {
 		flag.PrintDefaults()
 		return
 	}
-	file, err := os.Open(*imagePath)
-	if err != nil {
-		fmt.Printf("Could not open image file: %v\n", err)
-		os.Exit(1)
-	}
 
-	onProgress := func(progressMsg string) {
-		fmt.Println(progressMsg)
-	}
+	for i := 0; i < 10; i++ {
 
-	err = Update(context.Background(), file, deviceSerial, configuredAddr, onProgress)
+		file, err := os.Open(*imagePath)
+		if err != nil {
+			fmt.Printf("Could not open image file: %v\n", err)
+			os.Exit(1)
+		}
 
-	if err != nil {
-		fmt.Println(err.Error())
-		fmt.Println()
-		fmt.Println("Update failed. Try turning the Senso off and on, waiting for 30 seconds and then running this update tool again.")
-		os.Exit(1)
+		onProgress := func(progressMsg string) {
+			fmt.Println(progressMsg)
+		}
+
+		fmt.Printf("Update run %d out of %d\n", i+1, 10)
+		err = Update(context.Background(), file, deviceSerial, configuredAddr, onProgress)
+		fmt.Printf("Pausing for 5 second\n")
+		time.Sleep(5 * time.Second)
+
+		if err != nil {
+			fmt.Println(err.Error())
+			// fmt.Println()
+			// fmt.Println("Update failed. Try turning the Senso off and on, waiting for 30 seconds and then running this update tool again.")
+			// os.Exit(1)
+		}
 	}
 } 

[krksgbr]

There's one more detail I'm undecided about. What are your thoughts on this @knuton?

[knuton]

All very good, just some nitpicks on the progress messages, sorry! 🙃

I think the output is going to be easier to understand if we use uniform verbiage, such as "looking for", "found"/"not found", without also speaking about "discovery".

I believe the "to discover" verb just comes from mDNS/service discovery, and only makes sense when viewed from this technical angle. Otherwise it seems quite weird to me in English to say "I am trying to discover the lawnmower." or "I looked everywhere, but I could not rediscover my keys!".

Less importantly, would be nice to be consistent with whether or not messages end in periods.

[krksgbr]

Adding a couple more refinements to the batch, removing periods

[knuton]

🖖