[lldb-dev] LLDB Evolution: Next Phase

[Pavel Labath via lldb-dev]

On 19 August 2016 at 22:10, Kate Stone via lldb-dev
<lldb-dev at lists.llvm.org> wrote:
> Following up from yesterday, everything now builds and runs cleanly with the
> configuration described below on macOS.  I did have to make one set of minor
> changes to preserve #include ordering, but that’s already committed to trunk
> so there are no true blockers at this point.  To keep pushing forward I’d
> like to propose the following final schedule for comment:
>
> Today-Sept 4th Any final changes to prevent automatic formatting (see
> existing uses of // clang-format off and // clang-format on in source for
> reference.
> Sept 5th Trunk closes for commits while reformatting takes place and is
> validated before re-opening trunk.
>
> Following the reformatting trying to merge the resulting diff will be a
> nightmare for anyone with their own long-lived branch or fork of the
> project.  We’ll include a commit message describing the situation, but the
> sanest move is almost certainly to merge all diffs up to the formatting
> exercise and then simply format your own branch independently in lieu of
> actually attempting the merge.  We will ensure that nothing takes place in
> that commit aside from the automated formatting.
>
> We’ll be following the same strategy with the swift-lldb repository on
> GitHub.
>
> Any concerns?
>
> Kate Stone k8stone at apple.com
>  Xcode Low Level Tools

The plan sounds good to me. I have a question about the "validation"
part of the process.

Who will be doing that and what will be the exit criteria?

>From our side, we can consider the process a success if the following
three buildbots remain green after the reformat:
<http://lab.llvm.org:8011/builders/lldb-x86_64-ubuntu-14.04-buildserver>
(builds linux and android),
<http://lab.llvm.org:8011/builders/lldb-x86_64-ubuntu-14.04-cmake>
(build and test linux) and
<http://lab.llvm.org:8011/builders/lldb-windows7-android> (builds
windows). Note that the last buildbot also runs remote android tests,
which are still somewhat flaky, but the build step there is pretty
important as it using MSVC.

If anything goes wrong there, I will be available on Tuesday morning
(GMT) to diagnose and fix problems.

pl