Some fails:

brew install --build-from-source baidupcs-go
brew install --build-from-source charm
brew install --build-from-source dashing
brew install --build-from-source docker-swarm
brew install --build-from-source dvm
brew install --build-from-source gor
brew install --build-from-source perkeep
brew install --build-from-source source-to-image
brew install --build-from-source sync_gateway
brew install --build-from-source wellington
brew test --verbose aws-es-proxy
brew test --verbose cql
brew test --verbose dashing
brew test --verbose devd
brew test --verbose fn
brew test --verbose jabba
brew test --verbose mailhog

Looks like the repository of baidupcs-go vanished: https://github.com/iikira/BaiduPCS-Go.git

It doesn't seem to be a dependency of anything though. Maybe we should remove it @Homebrew/core ?

Looks like the repository of baidupcs-go vanished: https://github.com/iikira/BaiduPCS-Go.git

It doesn't seem to be a dependency of anything though. Maybe we should remove it @Homebrew/core ?

yeah, we should remove it, it is not rebuildable now.

@bayandin could you please rebase on top of latest master?

As the last build failed on Catalina (No space left on device) and we need to rerun it anyway, I've separated fixes for failed tests (which is unrelated to the current go upgrade) to a separate PR: #55844

@bayandin It's been building for more than 6h. Is this normal?

It is normal for Go.

For building Go on homebrew?

And testing all formulae which depend on Go to make sure it works with the new version.

~11 hours is normal for Go.

I see. Thank you!

@Bo98 is there anything we can do to help push this along?

Current failed steps:

brew test --verbose annie
brew install --build-from-source charm
brew test --verbose darksky-weather 
brew install --build-from-source dashing
brew test --verbose dashing
brew install --build-from-source dvm
brew test --verbose faas-cli
brew install --build-from-source gor
brew test --verbose jabba
brew test --verbose jabba
brew install --build-from-source perkeep
brew install --build-from-source source-to-image
brew install --build-from-source sync_gateway
brew install --build-from-source wellington

Ideally, these would all be resolved before merging. Some of the failures are coupled (e.g. dashing - test will fail since build failed).

@alebcay #52184

if you notice there I have opened PRs for perkeep and wellington. If the repos are sort of in maintenance mode could we not block merging at least for those?

@Bo98 if you have a list we cannot skip, I am happy to start working on trying to fix them

They finally cut a new release of charmstore-client, so if we update charm it should build with Go 1.14.

@Bo98 I feel like with every go release we come to packages that are broken and don't keep up.

What would a longer-term approach be here? use an internal version for homebrew downloads and builds and keep the external go keep moving along?

currently, charmstore has cut a release but quite a few other packages still have issues.

@Bo98 I feel like with every go release we come to packages that are broken and don't keep up.

What would a longer-term approach be here? use an internal version for homebrew downloads and builds and keep the external go keep moving along?

This is "rolling release package management" in a nutshell. It's not really avoidable.

The best way forward that wouldn't just help Homebrew is to help other projects improve their CI and code. That way we can avoid running into these issues altogether.

hmmm thanks for the info @MikeMcQuaid & @SMillerDev :)

I don't really understand too much of the internals. Very naively I was sort of hoping we could do something like use go@1.14.3 to build older breaking packages while moving go latest to be 1.14.4

But I also understand how that might become a nightmare with a lot of packages being left behind.

I'm also following this PR for some time and I think one of the improvements could be to cache correctly building packages. That way , if a package fails , we don't have to wait for 10 hours in the next run. I'm also aware that implementing this with Github actions is not trivial.

I'm sorry for not being around this week and following up here - I should have dealt with this pull request last week.

You are right that pre-existing failures are not a blocker for merging this.

I've gone through every single formula failing here and confirmed they are either already known as broken or not as a result of 1.14.4.

I will spend some time this weekend documenting these known broken formulae (or fixing the ones that can be fixed) so that it is more clearly known in the future.

@Bo98 has triggered a merge.

I will spend some time this weekend documenting these known broken formulae (or fixing the ones that can be fixed) so that it is more clearly known in the future.

Thanks @Bo98! Dumped the basics into #56539