issues and changes for release candidate 2

191 views
Skip to first unread message

Andrew Gerrand

unread,
May 6, 2013, 1:54:14 PM5/6/13
to golang-dev
Hi golang-dev,

We're gearing up to issue 1.1rc2.

In your opinions, what issues must be fixed, and what in-review changes must be submitted, before 1.1rc2 is cut? (Remember, this should be the final 1.1)

golang.org/issue/5337 is important, and has potential fix golang.org/cl/9226043

golang.org/issue/5402 seems important, and may have been misclassified as a dup of 5236.

golang.org/cl/9157044 is still under review and is important (whether we accept it or not)

golang.org/cl/8685045 is also pending, and seems minor enough that it should go in

Any others?

Andrew

Dave Cheney

unread,
May 6, 2013, 1:56:16 PM5/6/13
to Andrew Gerrand, golang-dev
There was a report that the tour shipped with windows still lacks a
.exe extension and/or exec bit,
> --
>
> ---
> You received this message because you are subscribed to the Google Groups
> "golang-dev" group.
> To unsubscribe from this group and stop receiving emails from it, send an
> email to golang-dev+...@googlegroups.com.
> For more options, visit https://groups.google.com/groups/opt_out.
>
>

Andrew Gerrand

unread,
May 6, 2013, 1:59:09 PM5/6/13
to Dave Cheney, golang-dev

On 6 May 2013 10:56, Dave Cheney <da...@cheney.net> wrote:
There was a report that the tour shipped with windows still lacks a
.exe extension and/or exec bit,

Nuts. Looks like I built go1.1rc1 with an old version of dist.exe.

Dave Cheney

unread,
May 6, 2013, 2:04:49 PM5/6/13
to Andrew Gerrand, golang-dev
Is it possible to make a branch or release tag for this next release candidate ?

Andrew Gerrand

unread,
May 6, 2013, 2:07:20 PM5/6/13
to Dave Cheney, golang-dev
On 6 May 2013 11:04, Dave Cheney <da...@cheney.net> wrote:
Is it possible to make a branch or release tag for this next release candidate ?

I'm happy to cut RC2 from a new release-branch.go1.1. I think the time has come.

Daniel Theophanes

unread,
May 6, 2013, 5:31:48 PM5/6/13
to golan...@googlegroups.com
I just read up on issue 5402 which is a dup of 5236. I don't care about having a fix to issue 5236 for Go1.1, but as I use Go on Windows at work, I'd be very interested to see the arena size (VM commit size?) lowered for windows. Right now my tip go program has reserved 142 TB of memory, while windows itself can handle only a fraction of that (32 GB) due to my SKU:

Thanks,
-Daniel

Andrew Gerrand

unread,
May 6, 2013, 5:40:32 PM5/6/13
to Daniel Theophanes, golang-dev

On 6 May 2013 14:31, Daniel Theophanes <kard...@gmail.com> wrote:
I just read up on issue 5402 which is a dup of 5236. I don't care about having a fix to issue 5236 for Go1.1, but as I use Go on Windows at work, I'd be very interested to see the arena size (VM commit size?) lowered for windows. Right now my tip go program has reserved 142 TB of memory, while windows itself can handle only a fraction of that (32 GB) due to my SKU:

But does it practically affect the operation of the program?

minux

unread,
May 6, 2013, 5:42:47 PM5/6/13
to Andrew Gerrand, golang-dev
On Tue, May 7, 2013 at 1:54 AM, Andrew Gerrand <a...@google.com> wrote:
We're gearing up to issue 1.1rc2.

In your opinions, what issues must be fixed, and what in-review changes must be submitted, before 1.1rc2 is cut? (Remember, this should be the final 1.1)

golang.org/issue/5337 is important, and has potential fix golang.org/cl/9226043

golang.org/issue/5402 seems important, and may have been misclassified as a dup of 5236.

golang.org/cl/9157044 is still under review and is important (whether we accept it or not)
CL 9157044 is in. 

golang.org/cl/8685045 is also pending, and seems minor enough that it should go in
how about this one?
https://codereview.appspot.com/9103045/ to enable the test TestExtraFilesFDShuffle in os/exec?

should we wait for the following CLs?
https://codereview.appspot.com/9164043/ better error message for cmd/go
https://codereview.appspot.com/9145044/ to fix newton method in doc/code.html: awaiting CLA
https://codereview.appspot.com/9149045/ for more platforms in cmd/api test?

Andrew Gerrand

unread,
May 6, 2013, 5:46:07 PM5/6/13
to minux, golang-dev
On 6 May 2013 14:42, minux <minu...@gmail.com> wrote:

On Tue, May 7, 2013 at 1:54 AM, Andrew Gerrand <a...@google.com> wrote:
We're gearing up to issue 1.1rc2.

In your opinions, what issues must be fixed, and what in-review changes must be submitted, before 1.1rc2 is cut? (Remember, this should be the final 1.1)

golang.org/issue/5337 is important, and has potential fix golang.org/cl/9226043

golang.org/issue/5402 seems important, and may have been misclassified as a dup of 5236.

golang.org/cl/9157044 is still under review and is important (whether we accept it or not)
CL 9157044 is in. 

golang.org/cl/8685045 is also pending, and seems minor enough that it should go in
how about this one?
https://codereview.appspot.com/9103045/ to enable the test TestExtraFilesFDShuffle in os/exec?

No
 
should we wait for the following CLs?
https://codereview.appspot.com/9164043/ better error message for cmd/go
 
No. This expands the API of go/build, which it's too late to do at this point.

https://codereview.appspot.com/9145044/ to fix newton method in doc/code.html: awaiting CLA

No
 
https://codereview.appspot.com/9149045/ for more platforms in cmd/api test?
 
It's reviewed, feel free to submit.

minux

unread,
May 6, 2013, 5:46:46 PM5/6/13
to Andrew Gerrand, Daniel Theophanes, golang-dev
the problem is: every Go program on windows, no matter how simple it is, will use 500MB+ of
committed memory (physical + page file), so we can't run more than 8 Go programs at the
same time on a machine with 4GB memory (physical + page file).

the good news is that doc/go1.1.html doesn't say how exactly Go 1.1's max heap size is
increased, so we can still change this "implementation detail" before Go 1.1 ships.

Daniel Theophanes

unread,
May 6, 2013, 5:48:34 PM5/6/13
to Andrew Gerrand, golang-dev
No. It will not affect operation of the program in question.
I just don't want it to take away VM from the system. I may not
entirely understand the issue, to be honest. But this doesn't make me
feel comfortable..

Andrew Gerrand

unread,
May 6, 2013, 5:51:54 PM5/6/13
to minux, Daniel Theophanes, golang-dev

On 6 May 2013 14:46, minux <minu...@gmail.com> wrote:
the problem is: every Go program on windows, no matter how simple it is, will use 500MB+ of
committed memory (physical + page file), so we can't run more than 8 Go programs at the
same time on a machine with 4GB memory (physical + page file).

the good news is that doc/go1.1.html doesn't say how exactly Go 1.1's max heap size is
increased, so we can still change this "implementation detail" before Go 1.1 ships.

Is the simplest fix to limit the arena size on Windows so that it doesn't reserve so much virtual memory?

minux

unread,
May 6, 2013, 5:56:05 PM5/6/13
to Andrew Gerrand, Daniel Theophanes, golang-dev
yes. i can propose a CL if you agree.

Andrew Gerrand

unread,
May 6, 2013, 5:59:14 PM5/6/13
to minux, Daniel Theophanes, golang-dev

On 6 May 2013 14:56, minux <minu...@gmail.com> wrote:
yes. i can propose a CL if you agree.

Okay, please send a CL.  I'm not sure if we want to go monkeying around with this stuff at this late stage, but more information can't hurt.

Thanks.

brainman

unread,
May 6, 2013, 10:23:40 PM5/6/13
to golan...@googlegroups.com
On Tuesday, 7 May 2013 03:54:14 UTC+10, Andrew Gerrand wrote:

> Any others?

I think golang.org/issue/5349 should be fixed. It affects plan9 and windows net.Dial for "udp" and even "tcp" on some systems. It is a race, so it might work sometimes and break next day. Mikio sent a fix https://codereview.appspot.com/8687045/, but he's been quiet lately, so I send a similar CL https://codereview.appspot.com/9159043/ too. Make a pick.

Alex
Reply all
Reply to author
Forward
0 new messages