Josh Bleecher Snyder

See also a slightly more general discussion of the interaction between concurrency and reproducibility guarantees at https://github.com/golang/go/issues/26263.

I just got bitten by this. Another option is to have vet check that MarshalJSON uses a value receiver.

The options for setting DNS are rather complicated. I think you may be able to accomplish your goals with Split DNS or a search domain. There are some docs at...

I’d say the former. I can’t think of a reason to prefer the string form. (I guess theoretically maybe if you have some weird string-interning-for-compression slog handler? Seems very rare...

> no real-world impact from race conditions in Fyne I can trigger a variety of crashes and race-y failures (mutex lock/unlock mismatches) with a couple of minutes of widget resizing....

> once the change stop, you eventually end up with consistency To achieve this "last writer wins" style of eventual consistency, you need to use the sync/atomic package for any...

The rendering thread calls user methods like Layout or MinSize, correct? And that user code can do anything at all, including writes. If user code is expected to provide its...

Here’s a panic backtrace: https://gist.github.com/josharian/3baab9809fdd6aca009a258a7916ad96 Here’s another: https://gist.github.com/josharian/ad747c4b3dc98b9d51096e306ac5c45a (note that I have no defers and no RWMutexes in my code.)

> One low hanging fruit is to go through the widgets and make sure that they do appropriate locking I started on this but found that the tests in `develop`...

OK, all the tests pass locally. My plan is to run `go test -v -race -failfast`, pick out the first failing test, fix it, send a PR, repeat. I'll have...