cc @phausler

The Darwin behavior is the expected format (that is what it does on macOS with the objc foundation). So we should update the linux code to emit the same values.

Ive just been looking at this. I updated the tests to work correctly on macOS and disabled them on linux for now. I think there is an underlying issue with ICU as the Foundation / CoreFoundation code is doing the right thing on Linux, its just the ICU isnt outputting the correct response. Although it seems to work correctly with a different timezone (PST instead of GMT) !

Comment by David Dunn (JIRA)

I've come to the same conclusion as you Simon Evans. I've also updated the MacOS tests, and currently looking at fixing the Linux tests which lead me to the incorrect version of ICU on Ubuntu (55.1 whereas 57.1 on MacOS).

We've been able to run the build script on Ubuntu 16.04 which picked up a newer version of ICU (57.1) using a change made in your PR Simon however this didn't fix the issue we were seeing in the tests. I've also been speaking to Tony Parker about this and he informed me that Apple actually ships its own additions to ICU as well and that these additions are used by Foundation in various places so it'd be that version of ICU he'd want shipping as a part of Swift. We haven't been able to get this particular version of ICU to work within the build script yet but I'll update with any progress we make.

Ive done a fix for the tests in #1147

Ive actually tested ICU59.1 directly with a small C program and its still broken with timezone GMT but actually works with timezone PST. It does however work on macOS using the system default icucore library. So I think it is actually an ICU issue not a linux foundation/corefoundation one