Jakob Borg

This surprises me as we have a test that verifies that we can remove read only directories in read only directories on Windows: https://github.com/syncthing/syncthing/blob/master/lib/osutil/osutil_test.go#L99 and we actually appear to be...

There's something else going on here beyond the read only bit. Read only directories can be deleted just fine. I tried to make a test for this, but if I...

Yeah I guess the only odd thing is the "access denied" vs the more reasonable "not empty" then. But this issue seems much more niche than I initially thought, and...

We don't and that's not the problem here. I also suspect the standard library fix (linked from the issue above) misses the point as it just does a chmod of...

https://github.com/syncthing/syncthing/blob/20aaa5927b982b7fdd01c173b58dc534f59aceee/lib/model/folder_sendrecv.go#L1826

I haven't tried but maybe an ignore pattern like `(?d)desktop.ini` would work (ignore desktop.ini and remove it if it's in the way of directory deletion).

We should look at what RemoveAll does under the hood to handle this and adopt the relevant parts. (We *want* the remove to fail if there are unexpected items in...

The "read only" part is a red herring, your chmod will do nothing to enable the removal of that directory. Removing desktop.ini should resolve it, but if I remember correctly...

Dammit, you guys. "Who cares" is not a good response to a filed issue; clearly at least @tomasz1986 did because he filed. "We don't track cosmetic issues" is clearly not...

Well, I mean, that’s a trade off that might be legitimate, so I don’t think we need to enforce it. There will presumably be some natural batching anyway for changes...