Hi, I am just wondering how “smartly designed” the database refreshes were for the v0.13->0.14 format. My understanding is that they were just deleted and fully recreated, no?
My setup is a Linux server, an OS X desktop, and a Windows desktop. My update process was: Linux server first in AM of 20th, then OS X at Noon of 20th, then Windows at Noon of 21st. I am trying to figure out if any of my data modified during this “not-everyone-on-same-version” timeframe could have caused stuff to be improperly modified.
Why I think this: a folder on my OS X client(the Master) was shown as out of sync on the 21st. This happened after I updated my Windows device. So I assume the database from Windows shows “these files were changed on July 21st” because of db refresh, even though I hadn’t modified these files since the evening of the 20th on OS X. If my OS X client wasn’t master, I think my data would have gotten (at least partially) overwritten by my out-of-sync Windows client, losing me a day’s worth of work. This worries me.
But maybe I am wrong and nothing would have happened? But based on the out-of-sync warning, I assume it would have messed up my previous day’s work
I know db recreations don’t need to be done usually, but how smart is it, exactly? What if someone has an old (and out of sync) device, decide to finally upgrade to v14, and then they have a total mess of data because of it?
Hopefully I am 100% wrong about the above situation being possible, because otherwise that seems a bit scary (especially for those ~2k people still using v12!). So thanks for any insight
PS: it sounds like V0.14 previously deleted files re-appear → ignore permission bits, but mine does not seem to relate to permissions. I also only use the base syncthing app, no wrappers.