Go 1.2 is released | Hacker News Reader

Go 1.2 is released

blog.golang.org

317 pointsenneff44 comments

Looks like an algorithmic or manual bitcoin-sweep of the front page took place. I welcome a programming language article to the #1 spot.

Official TLS 1.2 support? Oh yes please. Increases in JSON speed? Oh hell yes.

I'm loving how Go, as a package in the standard lib, now supports better TLS than most servers.

It seems kinda strange to me, honestly. All the big SSL libraries have had (and fixed) very subtle side-channel, timing, etc attacks... so why open ourselves to known exploits, if the fixes weren't re-implemented in Go?

They're implementing the reference so that the library can be improved over time. If you want a native implementation, you have to start somewhere.

The go TLS stack does need to be seriously audited, but that's expected, and it's going to take time for that to happen.

> re-implemented in Go

Conversely, this enables a host of features (e.g cross-compiling) unavailable or disabled were you to link against C-land using cgo.

That is a big potential issue, but really, you can deploy a standalone binary written in go within 2 minutes that supports the state of the art(minus perhaps some subtle but important fixes) in TLS.

If not anything else, that is pretty damn cool.

Like the other fellow said, it's not as though fixes to attacks will not be accepted into the language.

It's an open source language - surely critical security fixes are easily patched in?

I don't like Go's large number of inconsistencies. This includes the builtin function/keywords: len,cap,new,make,range and the built-in types: map,lists,arrays.

For example, as long as you were going to have built-in functions and keywords like len and cap, why not make them pretend to implement an interface so that user types could implement them?

Why can functions return multiple values but not channels?

Why vary range semantics between the different collection types? Seriously the first value over lists should be the index? Most everything else returns an error, but channels return ok for a second value?

They don't allow generic user types, but didn't even bother to adhere to some standard syntax for specifying the types in the built-in collections. I.E.: chan type, map[type]type, []type Sticking to some kind of system would at least give a preview of what user generics might have looked like. Now if they ever do implement generics they'll have to figure out some crazy system to make the built-in types make sense (or deprecate the old syntax).

There is a lot to like about Go, but I can't help but feel that re: hype, Go is the future MongoDB of PLs (overall I like MongoDB too).

Shameless plug: a few friends and myself ported the reference toolchain to Solaris (illumos). We hope to integrate this upstream now that Go 1.2 is released.

How'd you get past the split stack not working properly on SP... oh... I guess this means Illumos still isn't running on SPARC yet. ;)

Nice work BTW.

When I was involved with illumos development, it did run on SPARC. Don't know the status now. Not many people interested in keeping it alive, I guess.

I'd love the technical challenge in bringing Go to SPARC, but writing a new compiler is significant work. Gccgo works on SPARC, in the meantime.

1.3 they are planning to have contiguous stacks

Why wouldn't split stacks work on SPARC, is there something about the architecture that makes it impossible?

Is it Illumos specific or would it work on Solaris 10 as well?

It is not illumos-specific, but currently it does not work on Solaris 10. It would be trivial to make it work on Solaris 8-10. The main reason I haven't fixed it is because I don't have any machine running Solaris 8-10. We need a builder that does continuous integration.

If people are interested in these older systems, and if someone donates a builder, I can make Solaris 8-10 work as well.

We can provide you with one. How much cpu/mem do you need?

Would gcc 4.7.2 be ok? (I'd be grabbing any tools you needed from sunfreeware and that's the latest they have).

My email is in my profile, so we could move the discussion there.

Thanks for the offer, I can't find your email in your profile.

Odd. It's there but doesn't seem to show for others.

pauleyphonic @ that google email service.

People were also discussing this when the code hit, for those who didn't see that discussion.

https://news.ycombinator.com/item?id=6818371

This is slightly OT, but I figured this thread is a good pool of Go people =).

What books/resources would you guys recommend for a beginner to learn Go?

I'm coming from a (predominantly) Python background.

I've seen Mark Summerfield's book (he also wrote the the PyQT book and Programming in Python) on Go, which seems pretty good, and I think I'll take a shot at the official tutorial.

Is there anything else you guys would recommend?

Also, is it just me, or is the word "go" hard to search for - I know there's golang, but I'm always worried I'd miss things, if people just use "go" when they write.

http://tour.golang.org/ - as you already mentioned.

https://gobyexample.com/ - a nice resource for beginners.

http://learnxinyminutes.com/docs/go/ - quite dense but a good overview.

When I got started, I found Effective Go [0] and the Go Blog [1] extremely helpful. Effective Go goes over important properties of the language and Go Blog often go over important issues related to both language and standard libraries. I learnt about slices, interfaces, and json library all from Go Blog.

Also I think the best way to learn a language is to use it instead of reading a book. Pick up one of you old projects, and rewrite it in Go. Or pick up a new simple idea and implement in Go.

[0] http://golang.org/doc/effective_go.html [1] http://blog.golang.org/index

I initially had the same concern. "Golang" is a better keyword than you might think. Google is pretty smart to interpolate where it might be missing on certain pages.

Do the tour, read effective go, then start writing. As you have issues not solved by googling, join/ask golang nuts. Not a more helpful group of people in the tech world, in my opinion.

Here's the pull request for switching to 1.2 stable in Docker: https://github.com/dotcloud/docker/pull/2978

Have the GC issues been resolved yet (particularly on 32 bit)?

Not fully, but mostly. See comment 64 of issue 909. https://code.google.com/p/go/issues/detail?id=909

It's great that cgo now supports C++. I know that it's against Go's philosophy of short compile times, but more than once I had some C++ code that I liked to compile/link with Go code directly.

I think it's a good option to have, though some may see code compiled with Go that links to C++ code compiling slowly as the "supposedly fast" Go compilers fault.

This is perhaps a silly request: I would love to see a preview of the HTML output of this tool! Is there a page somewhere I could have a look at (really for curiosity more than anything else?)

http://i.imgur.com/PJBBMu8.png

What tool?

There's a tool that highlights the code to show which paths are covered by your test suite. (See the other reply for pic.)

I don't see `TextMarshaler` and `TextUnmarshaler` mentioned in `encoding/json` or `encoding/xml` document. Are there any reasons not doing that?

I think it's a documentation omission. Please file an issue: http://golang.org/issue/new

Thanks for the link. Will do that now

PR is up for the formula additions to homebrew.

And merged.

About test coverage. Is there techniques to exclude code that use the database so we get a higher percentage?

There is no mechanism for manually excluding code. But you can view the coverage profile in a web browser. We're going to publish a blog post about the new cover tool in a day or two.

Thanks!

The test and coverage features might bring me back to Go.

What made you decide to leave?

Possibly the lack of test and coverage features that might now bring him back to Go.

Reply on news.ycombinator.com