Subject: Re: [boost] [1.39.0]Release branch failures
From: Robert Ramey (ramey_at_[hidden])
Date: 2009-04-13 13:09:31
I would like to take this opportunity to restate some observations
and suggestions regarding the release process.
a) This process is much, much improved compared to that of
the previous one. 99% of the credit goes to Beman for getting
this going and sticking with it. And the boost library is significantly
larger than it was in the past so even this characterization doesn't
to justice to the scale of the accomplishment.
b) The current set of release failures on the release branch
are annoying but manageable. But still, there is room for
improvement at relatively little cost. I would propose that developers
do the following.
1) make the svn tree on your local machine based on the "release"
2) switch you're directories to the "trunk"
3) make your changes in your libraries and test on your machine.
and check into the trunk.
4) When you're satisfied, update your local boost. This will
make sure that you're local version reflects the next expected
5) Merge your set of changes into the release branch.
This should Maximize the chances that you come up with a clean release
Ideally, the trunk tests would be done in a similar way. That is, against
the last release except for the new check-ins. But until that happens,
until that happens, a large part of the benefit can be achieved by
an individual author by following the above plan. I've been using
it and it makes things much easier.
Beman Dawes wrote:
> Although most of the tests pass, default_toolset is failing on all
> platforms. There are a few other scattered failures.
> Please fix or markup,
> Unsubscribe & other changes:
Boost list run by bdawes at acm.org, gregod at cs.rpi.edu, cpdaniel at pacbell.net, john at johnmaddock.co.uk