No matter what I do right now, remove a device, add a folder, anything results in:
“Decoding posted config: unexpected EOF”
I have no idea what I did. I was adding some devices this morning and adding some folders to mutually sync that had already been syncing anyway. All of the sudden I just stopped being able to do anything.
Is it possible that adding a device manually from both sides simultaneously could cause some sort of race issue?
How do I even fix this though? Is the GUI somehow messing up the config it’s posting?
Actually, so weird, if I close my reverse proxy windows, then do any updates via direct connection, things work.
I’m pretty baffled. I think at this point the problem is due to the reverse proxy, but I have absolutely no idea what would be going on.
Why would this start suddenly? My suspicion is that old devices are sticking around in my config (mostly because I saw device names that are not in network anymore). And then my payload crossed some size threshold for nginx’s posts?
Just for confirmation, are you running the current version of Syncthing (which is v1.19.0 as of writing this post)?
You may want to post screenshots of the Web GUI and full logs taken after the error has popped up on the screen. Not an expert here, but I’d guess you may also want to enable config debug logging beforehand.
Also try a full refresh (Ctrl + F5 for example in Firefox) to reload the GUI. Sounds a bit like some other problems where the GUI state and code just got messed up in the browser.
Can you try with a different browser, or in a “private” browser window?
Oh god, I’m actually not running latest everywhere. My Linux boxes are on 1.18.6. I’m going to go fix that right now, so that I’m always tracking release. Or write a script to update it in nixpkgs or something so all NixOS users are updated regularly.
But anyway, if that doesn’t fix it - it’s at least somewhat reverse proxy related. I struggled for nearly 30 minutes. As soon as I opened 5x SSH Tunnels, closed my fancy reverse-proxied-URL tabs, and accessed via “localhost” instead of nginx, all of my issues went away (well, other than the issue where sometimes two machines keep thinking they need to sync changes when they don’t, but that usually requires some sort of unsync/resyncing and then it resolves itself.