Re: [boost] Fwd: Boost.Python build error in 1.58 RC1, 2 and 3 (OK in Beta 1)

On 09/04/15 11:40 AM, Beman Dawes wrote:

On Thu, Apr 9, 2015 at 10:01 AM, Olaf van der Spek wrote:

...

On Thu, Apr 9, 2015 at 3:51 PM, Marshall Clow wrote:

...

PLEASE STOP COMMITTING DIRECTLY TO MASTER!

That's not how we do development on Boost.

Commit your changes to the develop branch, give the testers a chance to run, and *then*, if everything seems OK, merge from develop to master. Can't this be handled by code / permissions?

Could you explain what you mean by "handled by code / permissions"?

While I understand how a continuous integration framework might only merge to master when develop tests are passing satisfactorily, I'm not aware of any way to prevent breaking merges with our current testing approach.

The entire issue is a result of me accepting a pull request to master, followed by a fix to an issue only observed on master. In addition I also wasn't fully aware of the use of master essentially as a release branch. While I will be more careful in the future to avoid such mistakes, I also think it might be useful to prevent (forbid) pull requests to master to begin with. Thanks, Stefan -- ...ich hab' noch einen Koffer in Berlin...