These themed days are inaverdent, but it happened again: Nearly everything done today had to do with encoding issues.
The big news is that it turned out everything written to files in the git-annex branch had unicode characters truncated to 8 bits. Now fixed so you should always get out the same thing you put in, no matter what encoding you use (but please use utf8). This affected things like storing repository descriptions, but worse, it affected metadata. (Also preferred content expressions, I suppose.)
With that fixed, there are still 7 source files left that use Char8 libraries. There used to be more; nearly every use of those is a bug. I looked over the remaining uses of it, and there might be a problem with Creds using it. I should probably make a push to stamp out all remaining uses of Char8.
Other encoding bugs were less reproducible.
And just now, Sören made some progress on Bootstrap3 icons missing on Android ... and my current theory is this is actually caused by an encoding issue too.