Daniel James wrote:

On 8 June 2014 10:22, John Maddock wrote:

...

The discussion on reducing dependencies is great, but folks, we do need to produce a release soon, or else it'll be back to bad old days of one release a year (if you're lucky). Just saying...

Marshall posted a possible schedule a few days ago, I put it on the calendar:

http://www.boost.org/development/

It was proposed that master closes for breaking changes on the 16th, which is a bit short notice now.

I, for one, have both seen Marshall's post and been keeping the release in mind. That said, the idea that modules need to freeze their master branch is a SVN relic. We are more evolved now; only the master branch of the superproject needs to freeze, and we can do that simply by stopping the script that updates the submodules on it.

Daniel James wrote:

On 8 June 2014 12:37, Peter Dimov wrote:

...

That said, the idea that modules need to freeze their master branch is a SVN relic. We are more evolved now; only the master branch of the superproject needs to freeze, and we can do that simply by stopping the script that updates the submodules on it.

We still need a way to include non-breaking changes from submodules.

Do you mean an automated way? If so, and if this requires modules to not update their master branches while a release is in progress... what was the point of the whole exercise?

On Sun, Jun 8, 2014 at 8:05 AM, Daniel James wrote:

...

Daniel James wrote:

...

On 8 June 2014 12:37, Peter Dimov wrote:

...

That said, the idea that modules need to freeze their master branch

is a

...

...

...

SVN relic. We are more evolved now; only the master branch of the > superproject needs to freeze, and we can do that simply by stopping

On 8 June 2014 12:55, Peter Dimov wrote: the >

...

...

...

script that updates the submodules on it.

We still need a way to include non-breaking changes from submodules.

Do you mean an automated way?

In whatever way works.

...

If so, and if this requires modules to not update their master branches while a release is in progress... what was the point of the whole exercise?

We change the process in the future if we want, but right now, we've got more than enough on our plates.

We don't have a viable release process in place yet, so we are going to have to make at least minimal changes. Straw man proposal: * On the 16th, (i.e. little more than a week from now) we turn off automatic sub-module master updates. * From that point until the release ships, a new library master release (with bug fixes only!) may be added manually to the Boost release by the release managers at their discretion. They have to be asked explicitly. * When the release ships, auto sub-module updates are turned back on. --Beman

On 8 Jun 2014 13:29, "John Maddock" wrote:

...

We don't have a viable release process in place yet, so we are going to have to make at least minimal changes.

Straw man proposal:

* On the 16th, (i.e. little more than a week from now) we turn off automatic sub-module master updates. * From that point until the release ships, a new library master release (with bug fixes only!) may be added manually to the Boost release by the release managers at their discretion. They have to be asked explicitly. * When the release ships, auto sub-module updates are turned back on.

That sounds sensible to me, John.

Why do you need to stop the submodule updates for master? Surely the correct thing to do is create a branch that targets the release.

On Sun, Jun 8, 2014 at 10:36 AM, Beman Dawes wrote:

On Sun, Jun 8, 2014 at 9:09 AM, James Sharpe wrote:

...

...

...

We don't have a viable release process in place yet, so we are going

to

...

have to make at least minimal changes.

Straw man proposal:

* On the 16th, (i.e. little more than a week from now) we turn off automatic sub-module master updates. * From that point until the release ships, a new library master release (with bug fixes only!) may be added manually to the Boost release by

On 8 Jun 2014 13:29, "John Maddock" wrote: the

...

...

release managers at their discretion. They have to be asked explicitly. * When the release ships, auto sub-module updates are turned back on.

That sounds sensible to me, John.

Why do you need to stop the submodule updates for master? Surely the correct thing to do is create a branch that targets the release.

Interesting question. Unfortunately, we want to keep regression testing on what will become the release, and don't have an automatic way to switch the current master testers to a different branch.

Or do we? Rene?

It's actually easy for the testers.. As all they have to do is change the "--tag=master" option to something else. The "hard" one is spinning up more resources to process the output and post test results. Also of note is that we likely don't have enough testing resources to run an additional branch. We would have to have the current master testers switch to something else. And the master results would then stop updating. -- -- Rene Rivera -- Grafik - Don't Assume Anything -- Robot Dreams - http://robot-dreams.net -- rrivera/acm.org (msn) - grafikrobot/aim,yahoo,skype,efnet,gmail

On 8 June 2014 13:19, Beman Dawes wrote:

We don't have a viable release process in place yet, so we are going to have to make at least minimal changes.

Straw man proposal:

* On the 16th, (i.e. little more than a week from now) we turn off automatic sub-module master updates. * From that point until the release ships, a new library master release (with bug fixes only!) may be added manually to the Boost release by the release managers at their discretion. They have to be asked explicitly. * When the release ships, auto sub-module updates are turned back on.

OK. Does every release manager have write access to the super project?

Paul A. Bristow wrote:

Ditto - *please* let's get a post modular-boost release out before we start any serious refactoring.

I did a double take when I read this, because it looks like you are recommending the opposite of what you're saying. But it's just that when you write "modular-boost", you're actually referring to non-modular-boost :).