[zorg] r214540 - Fixed mergefunc builder configure; added lldb builder for FreeBSD.

[David Blaikie]

On Mon, May 11, 2015 at 1:47 PM, Ed Maste <emaste at freebsd.org> wrote:

> On 8 May 2015 at 15:46, Ed Maste <emaste at freebsd.org> wrote:
> > On 8 May 2015 at 14:22, David Blaikie <dblaikie at gmail.com> wrote:
> >>
> >>
> >> Alternatively - could you just XFAIL all the current failures
> immediately
> >> (bringing the buildbot to green, at least), then work on fixing them?
> >
> > Unfortunately it's not that easy to XFAIL the remaining four because
> > they're being caught by a top-level timeout. I might have to just have
> > them skipped for now.
>
> Finally, we have one green build on the FreeBSD bot:
> http://lab.llvm.org:8011/builders/lldb-x86_64-freebsd/builds/5588
>
> For the sake of the archives, this is at r237010, and the build
> completed with one warning:
> /usr/bin/ld: warning: libedit.so.0, needed by
>
> /usr/home/buildslave/slave_as-bldslv5/lldb-x86_64-freebsd/llvm.obj/Release+Asserts/lib/liblldb.so,
> may conflict with libedit.so.7
>
> There's still flakiness in a handful of tests, at least when run on
> the bot (which I believe has a consistently high load, and is good at
> tripping over the races we have in LLDB).
>
> That said, I can't find a successful build in the available history for
> any of:
> lldb-x86_64-darwin-13.4
> lldb-x86_64-debian-clang
> lldb-x86_64-ubuntu-14.04-android
> lldb-x86_64-ubuntu-14.10
>
> While judging by eye lldb-x86_64-ubuntu-14.04-cmake seems to have
> about a 50% success rate.
>
> I'm trying to see about finding another FreeBSD LLDB buildslave I can
> add that would be less loaded, since the test suite passes pretty
> reliably for me on my own development machines on FreeBSD 10.x and
> 11.x. The flakiness and thread issues are real problems that will take
> a lot of effort to fix, but at least having a consistent, lightly
> loaded bot would help reduce the noise in ongoing development.
>

The buildbot could be configured to send you email when it fails without
sending blame email to everyone who committed the change. This might be the
right state until the flakiness is alleviated or fixed.

(it's still not great for the waterfall view of the buildbots ot have that
sort of thing - I'm not sure if we already have a grouping for buildbots
that are flaky/unreliable (if not, we should create one & pretty much put
all but a handful of the buildbots in there...)
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.llvm.org/pipermail/llvm-commits/attachments/20150511/f33b6e0a/attachment.html>