Override changes -- why?

I have the following setup.

System A is master (Linux). Systems B, C, and D are ordinary (client) nodes, one Linux and two Android. All systems are running the most recent verserion of SyncThing (v0.10.25 resp. v0.10.23). System A shares folder X to B, C and D. System B, C and D all share folder X back to A but NOT to each other. They have set the scan interval to zero since no local changes are expected. So far, so good. Every time when a change is made on system A to a file in folder X, the three clients start synching. The Androids display “Up to date” after a few seconds. The linux client however, jumps from “Syncing 100%” to “Syncing 0%” and the “Override changes” button shows on the master system. When I press it, it disappears but it reappears after a while.

What am I doing wrong?

Actually, I now know by heart the ticket number which relates to this.

Basically, we cannot modify mtime on Android, which causes Android to tell everyone his mtime. which causes to override masters version.

Once you click override, the master sends his mtime to everyone, but because Android fails to apply it, it sends it back as the new version, and the cycle continues.

This is only mtime though. Data should be fine.

Thanks for your reply. It sounds plausible, however:

[QUOTE]which causes Android to tell everyone his mtime[/QUOTE] In my setup the Android clients do not share to anyone (except the master). I fail to see how they can influence the linux client which they do not even know about.

Well it sends back something to the master which is newer than what the master has advertised, hence the master starts seeing the option to override the changes the Android device has broadcasted.

I would understand this if the ‘conflict’ was between the master and the mtime-incapable Android client. But it seems that the linux client gets stuck during the synch. The reappearing of the “Override changes” button may be, indeed, a master - Android client thing.

Anyway… Why can’t we set mtime on Android?

Well check the log on why the linux node is failing.

Regarding mtimes, - it’s all in the ticket.