Bugzilla

I have this caught in `rr` and have submitted it to pernosco, and I'm hoping maybe :ckerschb can take a look there and reconstruct what's going on.  (I can't provide a more direct response to comment 6's request-for-a-testcase, since (a) I don't know websockets particularly well, and (b) the homeassistant JavaScript is minified/obfuscated and has a lot of async hopping around.)

From skimming our C++, I suspect the relevant code here is in `WebSocketImpl::Init` which has a section:
```c++
  // If the HTTPS-Only mode is enabled, we need to upgrade the websocket
  // connection from ws:// to wss:// and mark it as secure.
```

I added a bit of logging (included in my rr/pernosco trace) which I confirmed does get triggered -- it's a good signal for when we're converting from `ws` to `wss` for the relevant URIs here.

That upgrade seems to be behind a call to `nsHTTPSOnlyUtils::ShouldUpgradeWebSocket`, which is returning `true` here, apparently because this if-check is failing:
```c++
  // 3. Check if NoUpgrade-flag is set in LoadInfo
  uint32_t httpsOnlyStatus = aLoadInfo->GetHttpsOnlyStatus();
  if (httpsOnlyStatus & nsILoadInfo::HTTPS_ONLY_EXEMPT) {
```

I have this caught in `rr` and have submitted it to pernosco, and I'm hoping maybe :ckerschb can take a look there and reconstruct what's going on.  (I can't provide a more direct response to comment 6's request-for-a-testcase, since (a) I don't know websockets particularly well, and (b) the homeassistant JavaScript is minified/obfuscated and has a lot of async hopping around.)

From skimming our C++, I suspect the relevant code here is in `WebSocketImpl::Init` which has a section:
```c++
  // If the HTTPS-Only mode is enabled, we need to upgrade the websocket
  // connection from ws:// to wss:// and mark it as secure.
```

I added a bit of logging (included in my rr/pernosco trace) which I confirmed does get triggered -- it's a good signal for when we're converting from `ws` to `wss` for the relevant URIs here.

That upgrade seems to be behind a call to `nsHTTPSOnlyUtils::ShouldUpgradeWebSocket`, which is returning `true` in this case, apparently because this if-check is failing:
```c++
  // 3. Check if NoUpgrade-flag is set in LoadInfo
  uint32_t httpsOnlyStatus = aLoadInfo->GetHttpsOnlyStatus();
  if (httpsOnlyStatus & nsILoadInfo::HTTPS_ONLY_EXEMPT) {
```

I have this caught in `rr` and have submitted it to pernosco, and I'm hoping maybe :ckerschb can take a look there and reconstruct what's going on.  (I can't provide a more direct response to comment 6's request-for-a-testcase, since (a) I don't know websockets particularly well, and (b) the homeassistant JavaScript is minified/obfuscated and has a lot of async hopping around.)

From skimming our C++, I suspect the relevant code here is in `WebSocketImpl::Init` which has a section:
```c++
  // If the HTTPS-Only mode is enabled, we need to upgrade the websocket
  // connection from ws:// to wss:// and mark it as secure.
```

I added a bit of logging (included in my rr/pernosco trace) which I confirmed does get triggered -- it's a good signal for when we're converting from `ws` to `wss` for the relevant URIs here.

That upgrade seems to be behind a call to `nsHTTPSOnlyUtils::ShouldUpgradeWebSocket`, which is returning `true` in this case, apparently because this if-check is failing -- so apparently this load isn't properly tagged as https-only-exempt:
```c++
  // 3. Check if NoUpgrade-flag is set in LoadInfo
  uint32_t httpsOnlyStatus = aLoadInfo->GetHttpsOnlyStatus();
  if (httpsOnlyStatus & nsILoadInfo::HTTPS_ONLY_EXEMPT) {
```