go1.1beta1 released

8,224 views
Skip to first unread message

Andrew Gerrand

unread,
Apr 4, 2013, 1:12:58 AM4/4/13
to golang-nuts
Hi Go nuts,

We have just released go1.1beta1, a test version of Go 1.1 cut from the default branch at revision 267bb9854177.

This is not a release candidate: We are still finalizing some details and tracking down a couple more bugs, but it should be ready for almost any purpose and there will be no significant changes other than bug fixes before the release. If you want to try out how Go 1.1 works for you, this is a way to get started early.

Please help us by testing your Go programs with the new tool chain and libraries, and report any problems using the issue tracker:
    http://golang.org/issue/new

You can download binary and source distributions from the usual place:
    https://code.google.com/p/go/downloads/list?q=go1.1beta1

To find out what has changed, read the release notes for Go 1.1:
    http://tip.golang.org/doc/go1.1html

Documentation for Go 1.1 is available at http://tip.golang.org/

Thanks,
Andrew

Paddy Foran

unread,
Apr 4, 2013, 1:18:17 AM4/4/13
to Andrew Gerrand, golang-nuts

Is that a typo? I got an error. http://tip.golang.org/doc/go1.1.html worked.

--
You received this message because you are subscribed to the Google Groups "golang-nuts" group.
To unsubscribe from this group and stop receiving emails from it, send an email to golang-nuts...@googlegroups.com.
For more options, visit https://groups.google.com/groups/opt_out.
 
 

Andrew Gerrand

unread,
Apr 4, 2013, 1:18:49 AM4/4/13
to Paddy Foran, golang-nuts
Yes, it's a typo. That's the correct link. Sorry.

Paddy Foran

unread,
Apr 4, 2013, 1:31:30 AM4/4/13
to Andrew Gerrand, golang-nuts

Hooray I didn't break everything!

Congratulations on the beta, Go team. This is really exciting.

Tony Shao

unread,
Apr 4, 2013, 1:56:59 AM4/4/13
to golan...@googlegroups.com
Great....

Tony Shao

unread,
Apr 4, 2013, 2:00:06 AM4/4/13
to golan...@googlegroups.com, Andrew Gerrand
http://tip.golang.org/doc/go1.1 this link is correct 

Tony Shao

unread,
Apr 4, 2013, 2:01:30 AM4/4/13
to golan...@googlegroups.com, Paddy Foran
hi, andrew , you miss a dot...

Eli Janssen

unread,
Apr 4, 2013, 2:23:10 AM4/4/13
to Tony Shao, golan...@googlegroups.com, Paddy Foran
Ran into an issue on linux-64bit (centos) using the supplied binary distrib tarball.
https://code.google.com/p/go/issues/detail?id=5203

No problems on darwin/OSX (built from source) so far.

On Apr 3, 2013, at 11:01 PM, Tony Shao <xio...@gmail.com> wrote:

> hi, andrew , you miss a dot...
> http://tip.golang.org/doc/go1.1.html
>
> On Thursday, April 4, 2013 1:18:49 PM UTC+8, Andrew Gerrand wrote:
>>
>> Yes, it's a typo. That's the correct link. Sorry.
>>
>>
>> On 4 April 2013 16:18, Paddy Foran <foran...@gmail.com <javascript:>>wrote:
>>
>>> Is that a typo? I got an error. http://tip.golang.org/doc/go1.1.htmlworked.
>>> On 4 Apr 2013 01:13, "Andrew Gerrand" <a...@google.com <javascript:>>
>>> wrote:
>>>
>>>> Hi Go nuts,
>>>>
>>>> We have just released go1.1beta1, a test version of Go 1.1 cut from the
>>>> default branch at revision 267bb9854177.
>>>>
>>>> This is not a release candidate: We are still finalizing some details
>>>> and tracking down a couple more bugs, but it should be ready for almost any
>>>> purpose and there will be no significant changes other than bug fixes
>>>> before the release. If you want to try out how Go 1.1 works for you, this
>>>> is a way to get started early.
>>>>
>>>> Please help us by testing your Go programs with the new tool chain and
>>>> libraries, and report any problems using the issue tracker:
>>>> http://golang.org/issue/new
>>>>
>>>> You can download binary and source distributions from the usual place:
>>>> https://code.google.com/p/go/downloads/list?q=go1.1beta1
>>>>
>>>> To find out what has changed, read the release notes for Go 1.1:
>>>> http://tip.golang.org/doc/go1.1html
>>>>
>>>> Documentation for Go 1.1 is available at http://tip.golang.org/
>>>>
>>>> Thanks,
>>>> Andrew
>>>>
>>>> --
>>>> You received this message because you are subscribed to the Google
>>>> Groups "golang-nuts" group.
>>>> To unsubscribe from this group and stop receiving emails from it, send
>>>> an email to golang-nuts...@googlegroups.com <javascript:>.

Phil Pennock

unread,
Apr 4, 2013, 3:37:25 AM4/4/13
to Andrew Gerrand, golang-nuts
On 2013-04-04 at 16:12 +1100, Andrew Gerrand wrote:
> We have just released go1.1beta1, a test version of Go 1.1 cut from the
> default branch at revision 267bb9854177.
>
> This is not a release candidate: We are still finalizing some details and
> tracking down a couple more bugs, but it should be ready for almost any
> purpose and there will be no significant changes other than bug fixes
> before the release. If you want to try out how Go 1.1 works for you, this
> is a way to get started early.

Sanity check, am not seeing mention in the release notes of the change
to reflect to be able to go into anonymous fields of a struct, so that
encoding/json and friends can show fields from embedded types?

Oversight, dropped, ... ?

Thanks for the great work, loving the performance improvements, amongst
other changes. :)

-Phil

Dan Kortschak

unread,
Apr 4, 2013, 3:54:17 AM4/4/13
to Phil Pennock, Andrew Gerrand, golang-nuts
xml can certainly deal with anonymous field, and the json documentation says it can.

Dave Cheney

unread,
Apr 4, 2013, 9:56:52 AM4/4/13
to Sanat Gersappa, golang-nuts
I'm sorry to hear that. By my estimates the compiler (not the linker)
uses 40-50% less memory and runs approx 1/3rd faster. Can you share
some details on your host.

On Thu, Apr 4, 2013 at 11:22 PM, Sanat Gersappa
<sanat.g...@gmail.com> wrote:
> Great work. Thanks. Looking forward to the final release.
>
> A small observation here - has the compiler slowed down since 1.0.3? - seems
> to take noticeably longer to compile stuff.
>
> - Sanat.

Sanat Gersappa

unread,
Apr 4, 2013, 11:04:13 AM4/4/13
to golan...@googlegroups.com, Sanat Gersappa
Ok. I tried building a simple helloworld program using go build. That was slow. 

However, on reading your reply, I tried compiling it with 8g and linking with 8l. Both were fast. So the problem could be 'go build' (and 'go run' as well). I'm on Windows 7 (32 bit) 2.40 Ghz i5 with 4GB RAM.

I just tried 'go run' and 'go build' on the same machine with 64-bit Ubuntu Linux and it is as blazingly fast as it was before. So it looks like a Windows-specific problem. Not sure if the same exists on 64-bit Windows.

Thanks.

Sanat.

John C.

unread,
Apr 4, 2013, 11:44:05 AM4/4/13
to golan...@googlegroups.com
When you say "noticeably slower", what time duration do you mean?  1s? 10s?

For the record, on my Win7-x64 laptop with go1.1beta1 my "go run" and "go build" times are similar to what they were on the same machine with go1.0.3, namely a little less than 1 second.  I do recall that when the go tool was introduced I was surprised that it took "so long" (~1s) whereas my batch files running 8g & 8l were quasi-instantaneous.  The difference between instantaneous and 1 second wasn't worth chasing down at the time and I never heard more about it.

DisposaBoy

unread,
Apr 4, 2013, 1:15:43 PM4/4/13
to golan...@googlegroups.com
How about go install? remember go build and go run will do the whole process from scratch every time

Karl J. Smith

unread,
Apr 4, 2013, 1:54:36 PM4/4/13
to DisposaBoy, golang-nuts
When go build seems slow to me, I do a 'go build -x -v'. If I see it compiling lots of included packages, I do a 'go install'.

The next go build will be faster as it won't be compiling those, too.




On Thu, Apr 4, 2013 at 10:15 AM, DisposaBoy <dispo...@dby.me> wrote:
How about go install? remember go build and go run will do the whole process from scratch every time

Guilherme Lino

unread,
Apr 4, 2013, 4:50:19 PM4/4/13
to golan...@googlegroups.com
why im i here? i should be doing something with my life..

Great work :)

Sanat Gersappa

unread,
Apr 4, 2013, 11:23:23 PM4/4/13
to golan...@googlegroups.com
Did another test. A cold start go build of helloworld took 11 seconds. But the next run took only 4 secs. So I guess it isn't worth agonizing over :-)

Brendan Tracey

unread,
Apr 5, 2013, 1:42:24 AM4/5/13
to golan...@googlegroups.com
I seem to remember that benchmarking does not report accurate results on OSX. Is that still the case?

Dave Cheney

unread,
Apr 5, 2013, 1:45:07 AM4/5/13
to Brendan Tracey, golang-nuts
Benchmarking reports accurate numbers on any platform which has a
functional clock; which is all of them :)

Profiling is less reliable on some versions of OS X, 10.6 especially
is considered unreliable.

David DENG

unread,
Apr 6, 2013, 11:14:41 AM4/6/13
to golan...@googlegroups.com
https://code.google.com/p/go/downloads/detail?name=go1.1beta1.darwin-amd64.tar.gz&can=2&q=go1.1beta1

$ uname -a
Darwin Davids-MacBook-Air.local 11.4.2 Darwin Kernel Version 11.4.2: Thu Aug 23 16:25:48 PDT 2012; root:xnu-1699.32.7~1/RELEASE_X86_64 x86_64

go run any-program:

go build command-line-arguments: signal: illegal instruction

FYI

David

Dean Sinaean

unread,
Apr 6, 2013, 10:58:58 PM4/6/13
to golan...@googlegroups.com
On 04/04/2013 01:12 AM, Andrew Gerrand wrote:
Hi Go nuts,

We have just released go1.1beta1, a test version of Go 1.1 cut from the default branch at revision 267bb9854177.

This is not a release candidate: We are still finalizing some details and tracking down a couple more bugs, but it should be ready for almost any purpose and there will be no significant changes other than bug fixes before the release. If you want to try out how Go 1.1 works for you, this is a way to get started early.

Please help us by testing your Go programs with the new tool chain and libraries, and report any problems using the issue tracker:
    http://golang.org/issue/new

You can download binary and source distributions from the usual place:
    https://code.google.com/p/go/downloads/list?q=go1.1beta1
I downloaded this beta, and unfortunately got an error(maybe the problem of my machine).
I got a cloud host which is running a ubuntu 11.04 x86_64. The error is like this:

wget http://go.googlecode.com/files/go1.1beta1.linux-amd64.tar.gz
tar xvzf go1.1beta1.linux-amd64.tar.gz
./go/bin/go
runtime: panic before malloc heap initialized
fatal error: runtime: cannot allocate heap metadata
Maybe the problem is about my cloud host, but I am not sure. Anyone got the same problem?

To find out what has changed, read the release notes for Go 1.1:
    http://tip.golang.org/doc/go1.1html

Documentation for Go 1.1 is available at http://tip.golang.org/

Thanks,
Andrew

Dmitry Vyukov

unread,
Apr 6, 2013, 11:12:54 PM4/6/13
to Dean Sinaean, golang-nuts
Go runtime reserves 256MB of virtual memory on startup, check ulimit -v.

peterGo

unread,
Apr 7, 2013, 8:47:33 AM4/7/13
to golan...@googlegroups.com
Andrew,

For the build, http://golang.org/ correctly displays: "Build version go1.0.3."

For the build, http://tip.golang.org/ incorrectly displays: "Build version devel +0a3866d6cc6b Mon Sep 24 20:08:05 2012 -0400."

Peter

Alexey Palazhchenko

unread,
Apr 7, 2013, 10:56:06 AM4/7/13
to golan...@googlegroups.com
Hi,


> Go runtime reserves 256MB of virtual memory on startup, check ulimit -v.

May it be configured at compile-time? I have the same problem during building from source on cheap VPS:
...
cmd/go

runtime: panic before malloc heap initialized
fatal error: runtime: cannot allocate heap metadata

–-–
Alexey "AlekSi" Palazhchenko

Sol Toure

unread,
Apr 7, 2013, 12:51:58 PM4/7/13
to Alexey Palazhchenko, golang-nuts
I get the same error on my laptop (Ubuntu 12.10/ Intel core i5 with 4GB of RAM).
I had a lot of windows opened, so closing some them made the problem go away.
However Go 1.0.3 works fine on the same machine with far more applications running.


krolaw

unread,
Apr 7, 2013, 2:13:19 PM4/7/13
to golan...@googlegroups.com
OpenVZ has been fixed for sometime, but many OVZ hosts are still default the older burst memory system.

If you are running Go, or pretty much anything, you'll want OVZ that uses vSwap. Even on a 64MB OVZ using vSwap I was able to run many Go apps. I also found rsync didn't run out of ram, when a 256MB with burst 512 vps would.

Therefore, when buying OZ vpses, always ask for vSwap. All of my hosts let me have vSwap for no extra charge.

Cheers.

Dmitry Vyukov

unread,
Apr 7, 2013, 4:45:16 PM4/7/13
to voidl...@gmail.com, golang-nuts
On Sun, Apr 7, 2013 at 10:00 AM, <voidl...@gmail.com> wrote:
The problem is some isolation containers have broken virtual memory accounting. Virtual memory is supposed to be (nearly) free; I'm amazed OpenVZ is still broken.

Yes, this is weird. What was they thinking about?..

We may try to use mmap(MAP_NORESERVE). But does it fix the problem on the VMs? MAP_NORESERVE allows to allocate more than RAM+swap, but one still needs ulimit -v = unlimited. So how they limit the amount of per-process virtual memory?




 
There are plenty of cheap VPSs that use tech with non-broken virtual memory accounting, for example KVM, Xen, LXC. One example I use for personal Go stuff: https://www.digitalocean.com/pricing

Andrew Gerrand

unread,
Apr 7, 2013, 7:12:52 PM4/7/13
to peterGo, golang-nuts

On 7 April 2013 22:47, peterGo <go.pe...@gmail.com> wrote:

For the build, http://tip.golang.org/ incorrectly displays: "Build version devel +0a3866d6cc6b Mon Sep 24 20:08:05 2012 -0400."

It's not actually incorrect; that's the version of godoc that runs tip.golang.org, rather than the version of the docs. Probably the string should be changed. I've filed a bug internally to address this.

Thanks,
Andrew

Dave Cheney

unread,
Apr 7, 2013, 7:29:29 PM4/7/13
to Sol Toure, Alexey Palazhchenko, golang-nuts
You haven't done something crazy like turn off swap or changed the VM overcommit settings?


Dean Sinaean

unread,
Apr 7, 2013, 8:51:44 PM4/7/13
to Alexey Palazhchenko, golang-nuts
On 04/07/2013 10:56 AM, Alexey Palazhchenko wrote:
> Hi,
>
> > Go runtime reserves 256MB of virtual memory on startup, check ulimit -v.
>
> May it be configured at compile-time? I have the same problem during
> building from source on cheap VPS:
Mine is the same problem. I've got only 512MB memory on the whole.
What's more, I used to run go program on a embedded ARM board with only
64MB memory. 256MB is too large for me.
> ...
> cmd/go
> runtime: panic before malloc heap initialized
> fatal error: runtime: cannot allocate heap metadata
>
> �-�
> Alexey "AlekSi" Palazhchenko

Dmitry Vyukov

unread,
Apr 7, 2013, 8:56:00 PM4/7/13
to Dean Sinaean, Alexey Palazhchenko, golang-nuts
On Sun, Apr 7, 2013 at 5:51 PM, Dean Sinaean <dean.s...@gmail.com> wrote:
> On 04/07/2013 10:56 AM, Alexey Palazhchenko wrote:
>>
>> Hi,
>>
>> > Go runtime reserves 256MB of virtual memory on startup, check ulimit -v.
>>
>> May it be configured at compile-time? I have the same problem during
>> building from source on cheap VPS:
>
> Mine is the same problem. I've got only 512MB memory on the whole. What's
> more, I used to run go program on a embedded ARM board with only 64MB
> memory. 256MB is too large for me.

You must be mixing physical RAM and process' virtual address space.
64-bit process on such a board still has 256TB of virtual address
space.

Dave Cheney

unread,
Apr 7, 2013, 8:56:27 PM4/7/13
to Dean Sinaean, Alexey Palazhchenko, golang-nuts
It is called _virtual_ memory for a reason. OpenVZ does some strange
accounting tricks, it continues to be a problem with Go (and Java)
runtimes.

Dean -- can you please open an issue if your 1.0.3 program will no
longer work under 1.1 (tip) on your arm host.

On Mon, Apr 8, 2013 at 10:51 AM, Dean Sinaean <dean.s...@gmail.com> wrote:
> On 04/07/2013 10:56 AM, Alexey Palazhchenko wrote:
>>
>> Hi,
>>
>> > Go runtime reserves 256MB of virtual memory on startup, check ulimit -v.
>>
>> May it be configured at compile-time? I have the same problem during
>> building from source on cheap VPS:
>
> Mine is the same problem. I've got only 512MB memory on the whole. What's
> more, I used to run go program on a embedded ARM board with only 64MB
> memory. 256MB is too large for me.
>
>> ...
>> cmd/go
>> runtime: panic before malloc heap initialized
>> fatal error: runtime: cannot allocate heap metadata
>>
>> –-–

Dean Sinaean

unread,
Apr 7, 2013, 8:57:09 PM4/7/13
to Dmitry Vyukov, golang-nuts
On 04/06/2013 11:12 PM, Dmitry Vyukov wrote:
Go runtime reserves 256MB of virtual memory on startup, check ulimit -v.
My machine returned "unlimited" using ulimit -v:
root@AY130304222943607e28:~# ulimit -v
unlimited

Dave Cheney

unread,
Apr 7, 2013, 8:59:38 PM4/7/13
to Dean Sinaean, Dmitry Vyukov, golang-nuts
If this is an OpenVZ host, please include the diff of your
/proc/user_beancounters file before and after (unsucessfully) running
your Go program. This will help us understand the limit which is being
breached.

Dmitry Vyukov

unread,
Apr 7, 2013, 9:17:20 PM4/7/13
to Dean Sinaean, golang-nuts
This is linux/amd64, right?
Try this patch https://codereview.appspot.com/8471048/

Dean Sinaean

unread,
Apr 7, 2013, 9:41:57 PM4/7/13
to Dmitry Vyukov, Dave Cheney, golang-nuts
On 04/07/2013 09:17 PM, Dmitry Vyukov wrote:
> This is linux/amd64, right?
> Try this patch https://codereview.appspot.com/8471048/
Right, the patch works. This makes a solution. Thank you all for your
kind help.

Dmitry Vyukov

unread,
Apr 7, 2013, 9:44:46 PM4/7/13
to Dean Sinaean, Dave Cheney, golang-nuts
On Sun, Apr 7, 2013 at 6:41 PM, Dean Sinaean <dean.s...@gmail.com> wrote:
> On 04/07/2013 09:17 PM, Dmitry Vyukov wrote:
>>
>> This is linux/amd64, right?
>> Try this patch https://codereview.appspot.com/8471048/
>
> Right, the patch works. This makes a solution. Thank you all for your kind
> help.


Yeah, but it is not committed.
Please file an issue at http://golang.org/issue/new

Dean Sinaean

unread,
Apr 7, 2013, 11:11:55 PM4/7/13
to Dmitry Vyukov, golang-nuts
On 04/07/2013 09:44 PM, Dmitry Vyukov wrote:
> On Sun, Apr 7, 2013 at 6:41 PM, Dean Sinaean <dean.s...@gmail.com> wrote:
>> On 04/07/2013 09:17 PM, Dmitry Vyukov wrote:
>>> This is linux/amd64, right?
>>> Try this patch https://codereview.appspot.com/8471048/
>> Right, the patch works. This makes a solution. Thank you all for your kind
>> help.
>
> Yeah, but it is not committed.
> Please file an issue at http://golang.org/issue/new
https://code.google.com/p/go/issues/detail?id=5236
That's it.

Alexey Palazhchenko

unread,
Apr 8, 2013, 3:51:13 AM4/8/13
to golang-nuts
Hi,

> You haven't done something crazy like turn off swap or changed the VM overcommit settings?

Dave, thank you very much. Swap wasn't enabled in default OS image I used. Enabling it solved my problem.

Re OpenVZ: I use KVM-based VPS, but it turned out it's possible to oversell it too.

Dmitry's comment at https://code.google.com/p/go/issues/detail?id=5236 scared me. If lazily heap allocation is out of scope for 1.1, I propose to not apply MAP_NORESERVE hack.

–-–
Alexey "AlekSi" Palazhchenko

Andrew Gerrand

unread,
Apr 8, 2013, 8:15:24 PM4/8/13
to Andrew Gerrand, golang-nuts

On 4 April 2013 18:37, Phil Pennock <phil-go...@spodhuis.org> wrote:
Sanity check, am not seeing mention in the release notes of the change
to reflect to be able to go into anonymous fields of a struct, so that
encoding/json and friends can show fields from embedded types?

What change?

Dan Kortschak

unread,
Apr 8, 2013, 9:20:02 PM4/8/13
to Andrew Gerrand, golang-nuts
He's referring to the change in encoding/{json,xml} to allow anonymous
fields to be marshaled and unmarshaled. Not a change in reflect.

Phil Pennock

unread,
Apr 8, 2013, 9:32:32 PM4/8/13
to Dan Kortschak, Andrew Gerrand, golang-nuts
On 2013-04-09 at 10:50 +0930, Dan Kortschak wrote:
> He's referring to the change in encoding/{json,xml} to allow anonymous
> fields to be marshaled and unmarshaled. Not a change in reflect.

Indeed, and sorry, I'd formed the mistaken impression that this had
required reflect-level support to happen.

Change a9fc9baa621b.

It's still not mentioned[*] in <http://tip.golang.org/doc/go1.1>.

Thanks,
-Phil

[*] I originally used the term "reflected" here, but further reflection
suggested that this might be a poor word choice.

Rob Pike

unread,
Apr 8, 2013, 10:49:00 PM4/8/13
to Dan Kortschak, Andrew Gerrand, golang-nuts
There were many changes made in since the last release. Those listed
in the document focus on API, language changes, and performance. I'd
rather not start listing all the bug fixes too.

-rob

Alexey Palazhchenko

unread,
Apr 9, 2013, 1:39:43 AM4/9/13
to Rob Pike, Dan Kortschak, Andrew Gerrand, golang-nuts
Hi,

> There were many changes made in since the last release. Those listed
> in the document focus on API, language changes, and performance. I'd
> rather not start listing all the bug fixes too.


Well, It's important behavior change, not a bug fix. Consider this example: http://play.golang.org/p/mwOdp6ONTm
Go 1.0 outputs {"S":"B"}
Go 1.1 outputs {} ... And this looks like a bug, actually.

–-–
Alexey "AlekSi" Palazhchenko

Rémy Oudompheng

unread,
Apr 9, 2013, 1:45:29 AM4/9/13
to Alexey Palazhchenko, Rob Pike, Dan Kortschak, Andrew Gerrand, golang-nuts
2013/4/9 Alexey Palazhchenko <alexey.pa...@gmail.com>:
> Hi,
>
>> There were many changes made in since the last release. Those listed
>> in the document focus on API, language changes, and performance. I'd
>> rather not start listing all the bug fixes too.
>
>
> Well, It's important behavior change, not a bug fix.

The Go 1.0 behaviour was documented as a bug to be fixed for a long time.

> Consider this example: http://play.golang.org/p/mwOdp6ONTm
> Go 1.0 outputs {"S":"B"}
> Go 1.1 outputs {} ... And this looks like a bug, actually.

Indeed, please file an issue.

Rémy.

Alexey Palazhchenko

unread,
Apr 9, 2013, 2:00:14 AM4/9/13
to Rémy Oudompheng, Rob Pike, Dan Kortschak, Andrew Gerrand, golang-nuts
Hi,

>> Well, It's important behavior change, not a bug fix.
>
> The Go 1.0 behaviour was documented as a bug to be fixed for a long time.

We may argue about definition of bug, but may I ask you to document this specific change at http://tip.golang.org/doc/go1.1? I feel like this will prevent a lot of WTFs from users.


>> Consider this example: http://play.golang.org/p/mwOdp6ONTm
>> Go 1.0 outputs {"S":"B"}
>> Go 1.1 outputs {} ... And this looks like a bug, actually.
>
> Indeed, please file an issue.

https://code.google.com/p/go/issues/detail?id=5245

Thank you.

–-–
Alexey "AlekSi" Palazhchenko

Dan Kortschak

unread,
Apr 9, 2013, 2:06:39 AM4/9/13
to Alexey Palazhchenko, Rémy Oudompheng, Rob Pike, Andrew Gerrand, golang-nuts
It is marked pretty clearly in the respective package docs and has been
asked about (the possibility of a fix) for quite some time here, so I'd
be surprised if most people either didn't already know (those who cared
about it before) or were not aware that it was any different to the
Go1.1 situation (people who didn't use it before).

Rémy Oudompheng

unread,
Apr 9, 2013, 2:13:04 AM4/9/13
to Alexey Palazhchenko, Rob Pike, Dan Kortschak, Andrew Gerrand, golang-nuts
On 2013/4/9 Alexey Palazhchenko <alexey.pa...@gmail.com> wrote:
> Hi,
>
>>> Well, It's important behavior change, not a bug fix.
>>
>> The Go 1.0 behaviour was documented as a bug to be fixed for a long time.
>
> We may argue about definition of bug,

It is documented as such in Go 1.0 documentation:
http://golang.org/pkg/encoding/json/#pkg-note-BUG

Rémy.

Job van der Zwan

unread,
Apr 9, 2013, 6:03:43 AM4/9/13
to golan...@googlegroups.com, Alexey Palazhchenko, Rémy Oudompheng, Rob Pike, Andrew Gerrand
Repeated questions on the go-nuts board does not mean everyone knows about it by now - it's not like every Go user reads every post here. And you can't assume that people will re-read all of the relevant package docs to see if there are changes to the packages they use. That's what a release document is supposed to be for: listing all the relevant changes. This is mentioned in the package docs:

> Handling of anonymous struct fields is new in Go 1.1. Prior to Go 1.1, anonymous struct fields were ignored. To force ignoring of an anonymous struct field in both current and earlier versions, give the field a JSON tag of "-".

In my opinion, not mentioning this in the minor changes list of the release notes would give a false first impression that this didn't change.

I don't see the need to list all bug fixes, that will just turn into noise for most readers, but not all bug fixes are created equal.

Brendan Tracey

unread,
Apr 10, 2013, 2:36:08 AM4/10/13
to golan...@googlegroups.com, Alexey Palazhchenko, Rémy Oudompheng, Rob Pike, Andrew Gerrand
The downloads at the link at the top of the thread are now listed as depricated. I assume it is better to use the go1.1beta at https://code.google.com/p/go/downloads/list ?

Thanks

Jeff R. Allen

unread,
Apr 10, 2013, 3:49:36 AM4/10/13
to golan...@googlegroups.com, Alexey Palazhchenko, Rémy Oudompheng, Rob Pike, Andrew Gerrand

On Wednesday, April 10, 2013 8:36:08 AM UTC+2, Brendan Tracey wrote:
The downloads at the link at the top of the thread are now listed as depricated. I assume it is better to use the go1.1beta at https://code.google.com/p/go/downloads/list ?

Go1.1beta1 is deprecated because Go1.1beta2 is out, and listed at the bottom of that page. The difference I noticed is that beta 2 on Linux will work with libc before 2.14, like on my Debian machine.

Thanks for fixing that, Andrew!

 -jeff 

Dave Cheney

unread,
Apr 10, 2013, 6:48:10 AM4/10/13
to Jeff R. Allen, golan...@googlegroups.com, Alexey Palazhchenko, Rémy Oudompheng, Rob Pike, Andrew Gerrand
That is because beta 2 was built on Debian stable :)
--

minux

unread,
Apr 10, 2013, 10:30:20 AM4/10/13
to Joel Neely, golan...@googlegroups.com
On Wed, Apr 10, 2013 at 8:16 PM, Joel Neely <joel....@gmail.com> wrote:
> TL;DR: My attempts to install 1.1 (beta2) have failed with installer errors
> as shown below. I'll be glad to provide any additional information that will
> help resolve this.
>
> Thanks!
> -jn-
>
> I downloaded beta2 for darwin/amd64 and attempted to install on my MacBook
> Pro running Mountain Lion (10.8).
>
> I got the warning: "A previous installation of Go exists at /usr/local/go.
> This installer will remove the previous installation prior to installing.
> Please back up any data before proceeding.", but continued the installation,
> having just done a full backup.
>
> After the usual "barometer" display indicating progress, installation failed
> with the message "The Installer encountered an error that caused the
> installation to fail. Contact the software manufacturer for assistance."
> displayed normally inside the installer dialogue. There was an additional
> message, in smaller, bold-faced text at the top of the dialogue that read:
> "There were errors with the installation. You may want to try installing
> again."
>
> Reattempts (including after renaming the existing /usr/local/go to
> /usr/local/go1) yielded the same result.
>
> AFAICT, the relevant portion of install.log reads as follows:
>
> Apr 10 07:04:47 Joels-MacBook-Pro.local installd[601]: PackageKit: Executing
> script "./preinstall" in
> /private/tmp/PKInstallSandbox.h4hXoj/Scripts/com.googlecode.go.WnC7Kl
> Apr 10 07:04:47 Joels-MacBook-Pro.local installd[601]: PackageKit: ***
> launch path not accessible
> Apr 10 07:04:47 Joels-MacBook-Pro.local install_monitor[607]: Temporarily
> excluding: /Applications, /Developer, /Library, /System, /bin, /private,
> /sbin, /usr
> Apr 10 07:04:48 Joels-MacBook-Pro.local install_monitor[607]: Re-included:
> /Applications, /Developer, /Library, /System, /bin, /private, /sbin, /usr
> Apr 10 07:04:48 Joels-MacBook-Pro.local installd[601]: PackageKit: releasing
> backupd
> Apr 10 07:04:48 Joels-MacBook-Pro.local installd[601]: PackageKit: allow
> user idle system sleep
> Apr 10 07:04:50 Joels-MacBook-Pro.local installd[601]: PackageKit: Install
> Failed: Error Domain=PKInstallErrorDomain Code=112 "An error occurred while
> running scripts from the package “go1.1beta2.darwin-amd64.pkg”."
> UserInfo=0x7ffa843024e0 {NSFilePath=./preinstall,
> NSURL=file://localhost/Users/joel/Downloads/go1.1beta2.darwin-amd64.pkg#com.googlecode.go.pkg,
> PKInstallPackageIdentifier=com.googlecode.go, NSLocalizedDescription=An
> error occurred while running scripts from the package
> “go1.1beta2.darwin-amd64.pkg”.} {
> NSFilePath = "./preinstall";
> NSLocalizedDescription = "An error occurred while running scripts from
> the package \U201cgo1.1beta2.darwin-amd64.pkg\U201d.";
> NSURL =
> "file://localhost/Users/joel/Downloads/go1.1beta2.darwin-amd64.pkg#com.googlecode.go.pkg";
> PKInstallPackageIdentifier = "com.googlecode.go";
> }
This is: https://code.google.com/p/go/issues/detail?id=5258

Sugu Sougoumarane

unread,
Apr 11, 2013, 5:40:44 PM4/11/13
to golan...@googlegroups.com
Is there an hg tag for this, or should we just use tip?

Rob Pike

unread,
Apr 11, 2013, 6:08:26 PM4/11/13
to Sugu Sougoumarane, golan...@googlegroups.com
I suggest using tip. The beta is really about the packaged downloads.

-rob

On Thu, Apr 11, 2013 at 2:40 PM, Sugu Sougoumarane <sou...@google.com> wrote:
> Is there an hg tag for this, or should we just use tip?
>
Reply all
Reply to author
Forward
0 new messages