go 1.5.2 timeline

[Michael Hudson-Doyle]

Hello,

I was wondering about the timeline, if there is any, for the 1.5.2
release. The background is that we're going to upload at least one fix
to Ubuntu soon, but if 1.5.2 is around the corner, I'd probably wait
until that was released instead.

Cheers,
mwh

[Andrew Gerrand]

There are no plans to do 1.5.2 any time soon, as there is no issue severe enough to trigger it right now.

​

[Dave Cheney]

I think https://github.com/golang/go/issues/13160 should deserve a
1.5.2 release.

On Fri, Nov 13, 2015 at 10:08 AM, Andrew Gerrand <a...@golang.org> wrote:
> There are no plans to do 1.5.2 any time soon, as there is no issue severe
> enough to trigger it right now.
>

> --
> 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/d/optout.

[Andrew Gerrand]

13160's worst case scenario seems to be unnecessary memory retention on 386 systems.

Doesn't seem severe enough to warrant a point release.

[Dave Cheney]

I think it could be more serious than that.
https://github.com/golang/go/issues/13176#issuecomment-156167822

[Caleb Spare]

13160's worst case scenario seems to be unnecessary memory retention on 386 systems.

Doesn't seem severe enough to warrant a point release.

​In https://github.com/golang/go/issues/13176, which is a runtime crash bug, Austin said he suspects it's the same as 13160.​

 

[Ian Lance Taylor]

I'm concerned by 13176 and 13143. I think 13160 is more serious than
you are describing: it leads to unexpected program crashes.

Ian

[Shane Hansen]

I've ran into problems caused by https://github.com/golang/go/issues/12932 during the last couple of days. Basically I'm unable to profile a small application because the runtime crashes hard. It might be worth including in a go1.5.2 release.

[Russ Cox]

+austin

It's true that there are no plans yet, but I think it's also true that we need to start planning. As I understand it, there have been a few serious runtime bugs that may warrant issuing 1.5.2. Austin can say better what they are and how critical they are.

Russ

[Austin Clements]

You beat me to the punch. I was going to suggest that we start thinking about a 1.5.2.

Here are some bugs I think warrant it:

https://golang.org/issue/13160, https://golang.org/issue/12552, and probably https://golang.org/issue/13176 are all related and lead to runtime crashes because of partial pointer writes.

https://golang.org/issue/12932 leads to random runtime crashes when profiling is enabled.

https://golang.org/issue/12894 can lead to significant performance problems. We've cherry-picked this one internally because of the grief it was causing.

https://golang.org/issue/12498 causes deadlocks in programs that call C that uses cancellations or setuid/setgid.

https://golang.org/issue/12233 causes programs with large heaps to often crash.

There's at least one project that's rolled back to 1.4 because of some of these issues. And we've cherry-picked the fixes for some of these internally at Google (12233 and 12894 at least), so they're clearly important to Google.

There are still open 1.5.2 bugs, some of which I'd like to fix (I'm working on 13143 right now and may have a fix), but they all seem incredibly rare or lower priority.

[ja...@nym.se]

Den fredag 13 november 2015 kl. 00:09:09 UTC+1 skrev Andrew Gerrand:

There are no plans to do 1.5.2 any time soon, as there is no issue severe enough to trigger it right now.

At least Syncthing is locked on 1.4 until https://github.com/golang/go/issues/12301 is released. It breaks anything trying to do multicast on Windows.

//jb 

[Brian Hatfield]

This seems like an important and valuable set of fixes to release.

Like Google, we've also cherry-picked 12233 in our environment, but I'd really rather stay on mainstream versions of Go as much as possible.

A silly question, but why not roll these up as 1.5.2 and release further fixes when they are ready as 1.5.3? Tracking some of these issues, it seems like folks would really appreciate it.

Respectfully,
Brian

[David Symonds]

On 16 November 2015 at 01:17, Brian Hatfield <bmhat...@gmail.com> wrote:

> A silly question, but why not roll these up as 1.5.2 and release further
> fixes when they are ready as 1.5.3? Tracking some of these issues, it seems
> like folks would really appreciate it.

There's a fair bit of work that goes into a release, so it's a
balancing act between getting point releases out quickly enough and
doing an unnecessary number of point releases when we could have
waited a little longer.

[Michael Hudson-Doyle]

On 14 November 2015 at 04:28, Austin Clements <aus...@google.com> wrote:
> You beat me to the punch. I was going to suggest that we start thinking
> about a 1.5.2.

\o/

> Here are some bugs I think warrant it:
>
> https://golang.org/issue/13160, https://golang.org/issue/12552, and probably
> https://golang.org/issue/13176 are all related and lead to runtime crashes
> because of partial pointer writes.
> https://golang.org/issue/12932 leads to random runtime crashes when
> profiling is enabled.
> https://golang.org/issue/12894 can lead to significant performance problems.
> We've cherry-picked this one internally because of the grief it was causing.
> https://golang.org/issue/12498 causes deadlocks in programs that call C that
> uses cancellations or setuid/setgid.
> https://golang.org/issue/12233 causes programs with large heaps to often
> crash.

I'm mostly interested in the non-atomic-pointer writes ones and a
couple of ppc64 codegen bugs.

Also, there is the issue that building Go with a very new C toolchain
fails (https://github.com/golang/go/issues/13114) -- is it appropriate
to fix this in a point release? I'll need to include this patch in any
update of Go in Ubuntu in any case...

> There's at least one project that's rolled back to 1.4 because of some of
> these issues. And we've cherry-picked the fixes for some of these internally
> at Google (12233 and 12894 at least), so they're clearly important to
> Google.
>
> There are still open 1.5.2 bugs, some of which I'd like to fix (I'm working
> on 13143 right now and may have a fix), but they all seem incredibly rare or
> lower priority.

Is there anything I can do to help?

Cheers,
mwh

[Austin Clements]

On Sun, Nov 15, 2015 at 7:20 PM, Michael Hudson-Doyle <michael...@canonical.com> wrote:

On 14 November 2015 at 04:28, Austin Clements <aus...@google.com> wrote:
> You beat me to the punch. I was going to suggest that we start thinking
> about a 1.5.2.

\o/

> Here are some bugs I think warrant it:
>
> https://golang.org/issue/13160, https://golang.org/issue/12552, and probably
> https://golang.org/issue/13176 are all related and lead to runtime crashes
> because of partial pointer writes.
> https://golang.org/issue/12932 leads to random runtime crashes when
> profiling is enabled.
> https://golang.org/issue/12894 can lead to significant performance problems.
> We've cherry-picked this one internally because of the grief it was causing.
> https://golang.org/issue/12498 causes deadlocks in programs that call C that
> uses cancellations or setuid/setgid.
> https://golang.org/issue/12233 causes programs with large heaps to often
> crash.

I'm mostly interested in the non-atomic-pointer writes ones and a
couple of ppc64 codegen bugs.

I see a few closed ppc64 1.5.2 issues. Are there any fixes you wanted that weren't marked 1.5.2?

Also, there is the issue that building Go with a very new C toolchain
fails (https://github.com/golang/go/issues/13114) -- is it appropriate
to fix this in a point release? I'll need to include this patch in any
update of Go in Ubuntu in any case...

It looks like Ian added to the 1.5.2 milestone. So, apparently yes. :)

> There's at least one project that's rolled back to 1.4 because of some of
> these issues. And we've cherry-picked the fixes for some of these internally
> at Google (12233 and 12894 at least), so they're clearly important to
> Google.
>
> There are still open 1.5.2 bugs, some of which I'd like to fix (I'm working
> on 13143 right now and may have a fix), but they all seem incredibly rare or
> lower priority.

Is there anything I can do to help?

https://github.com/golang/go/milestones/Go1.5.2 :)

It looks like there are a few that could use your expertise.

[Michael Hudson-Doyle]

I don't think so. There's https://github.com/golang/go/issues/12108
but that doesn't cause problems in 1.5 because duffzero is disabled.

>> Also, there is the issue that building Go with a very new C toolchain
>> fails (https://github.com/golang/go/issues/13114) -- is it appropriate
>> to fix this in a point release? I'll need to include this patch in any
>> update of Go in Ubuntu in any case...
>
>
> It looks like Ian added to the 1.5.2 milestone. So, apparently yes. :)

yay.

>> > There's at least one project that's rolled back to 1.4 because of some
>> > of
>> > these issues. And we've cherry-picked the fixes for some of these
>> > internally
>> > at Google (12233 and 12894 at least), so they're clearly important to
>> > Google.
>> >
>> > There are still open 1.5.2 bugs, some of which I'd like to fix (I'm
>> > working
>> > on 13143 right now and may have a fix), but they all seem incredibly
>> > rare or
>> > lower priority.
>>
>> Is there anything I can do to help?
>
>
> https://github.com/golang/go/milestones/Go1.5.2 :)
>
> It looks like there are a few that could use your expertise.

I made a few comments. The compiler code and I are not friends :-)

Cheers,
mwh