Motiejus Jakštys Public Record

My Zig and Go work for the next 3 months


Now that Uber uses “zig cc” to compile it’s Go Monorepo, here is a sneak preview of what I have planned for the next 3 months on the Zig/Go side:

  • Finish an “official” Uber’s blog post about how Uber uses zig and why. The content will be similar to my post a year ago, but with the milestones closed. Since the previous post we have enabled zig-cc as the default C/C++ compiler for CI, dev and production, our toolchain got open sourced and the company has committed to maintaining it. We still find bugs in Zig and Go and interaction between both though, so there is still work to be done.
  • Play around with MacOS target support for the mentioned toolchain. I can’t promise anything, but at least I want to evaluate the feasibility. This needs a bit of ground work on the toolchain side plus a lot of testing.
  • Make all Go tests pass with CC="zig cc". This is a prerequisite for the next goal.
  • Propose and, if accepted, implement an official CI node that tests Go with “zig cc” as the C compiler for linux/amd64 and linux/arm64. When this is done, the Go team will be alerted of regressions that impact “zig cc” users during the change. At the moment, unfortunately, if Go has a regression that impacts “zig cc”, the Go team learns about it only when someone (we?) attempt to upgrade Go to the latest release.
  • Extension from the above: do the same for macos. This would have caught this issue way before the release of Go 1.20; it will now be fixed only in Go 1.20.3.

If you are my (ex-)colleague that is/was involved, thank you very much. We have such an amazing team that was able to pull this off.