Web disconnections with 2021r3

Yes, it affects Xojo Cloud.

Yes, just had the disconnect on my iPad testing https://66679.xojocloud.net/

To reproduce: go to the website click OK scroll to the very last name ( Byron Yates ) click the name, and wait about a minute. That seems to trigger the bug fairly reliably for me.

Also, I noticed that when I reload the webpage, the picture for Byron Yates does not show up anymore. Related bug? separate issue?

Visited that URL and got a disconnect. Added a video to
66779 - Web disconnections with 2021r3

It took around 1 min 14 seconds after reloading the page for it to disconnect.

Yep, 1 min and 10 or seconds happens to my webapp also. I’ve timed it and describe it in my other post about this… hope now a QUICK FIX/PATCH will come and not have to wait around for the next release, because this bug makes webapps unusable. well, serious webapps, not 2 controls thrown together.

An example has been provided on Xojo Cloud with 2021r3.

Interesting on the same computer (Windows 11) the problem occurs much more often under Chrome (96.0.4664.45) as under Microsoft Edge (Chromium) 96.0.1054.34 … but it happens there too, not every minute like in Chrome, maybe every 10 minutes under Edge.

It also happens with Safari on my M1

Take a look at the serverevent and ping XHR requests in a debugger, you’ll notice that they are all sent with Keep-Alive (which makes sense) but the timeout settings are different for each one:

timeout=5, max=100
timeout=5, max=99
timeout=5, max=98
timeout=5, max=97
timeout=5, max=95
timeout=5, max=96

Per Keep-Alive - HTTP | MDN the max setting is the maximum # of connections that can happen over this connection before closing it.

  • Why would this number be counting down?
  • is it a coincindence that this number is under 100 and most people are seeing disconnects roughly at 100 seconds?

@Andreas_Reichmann could you put up the 2012R2.1 version on a different server URL? It would be interesting to compare the behavior of the two.

Here it is build with 2021 R2.1 … works without problems

Build with R3
https://66679.xojocloud.net … connection fails

Also reproducible with a server running Windows. As soon as access is made via any proxy (tested with Apache, Nginx, HaProxy) the connection breaks.

(Edit: Info about missing maps key).
And keep in mind … i have no valid maps key in my uploaded samples


Having same issue with WebApps. I need proxy for speed and load balancing, but so far only stand alone WebApps do not display this behavior, with an exception. It’s not every app though. Found that one of my other apps which requires the old /API/ for handlespecialurl does not display this. The only difference between this app and base WebApps is that I built a drop-in wrapper to continue to implement HandleSpecialUrl in new WebApps (I needed the path, handleurl was not an option). I built an empty app and dropped my API URL path handler, and it’s been running ever since without a crash on my server. I don’t see how the path would affect anything since no resources are being requested by xojo for the path, but the app continues to run without down time through my Nginx/proxy setup. Hopefully this issue can be tracked down.

1 Like

@Andreas_Reichmann : can you run both versions with the –logging option enabled? Do you see anything specific happening when the R3 version loses connection? Also, does the proxy server itself have any event or error logging files that you can view?

I think we have reached the point where Xojo should be able to see and fix the problem. I have no free time next days to play around the bug.


You can see everything I posted on the forum (logs, console errors, etc…). I don’t think you can do anything else being a xojo issue and not a configuration issue.


5 or 6 days have passed and still no official word about this mega-issue… nice


@Gabriel_L, @Andreas_Reichmann raised feedback case 66779 which is now marked as fixed. Therefore I assume the “fix” will be applied in the next release. Given, that the issue is a “show-stopper” for Web 2.0 deployment, a reasonable person might assume a point release 2021 R3.1 might be on the near horizon (then again it might not). I hope that helps.

Kind regards, Andrew

1 Like

A bug of this magnitude that renders pointless the tool, I think it deserves some acknowledgment and if it was fixed, a quick patch release, not needing to wait for a next release that could happen in a few months. Operating on maybes and assumptions helps absolutely no one.