[lldb-dev] A commit between r227595 and r228444 broke tests on FreeBSD

[Ed Maste]

On 13 February 2015 at 20:49, Ed Maste <emaste at freebsd.org> wrote:
> On 13 February 2015 at 20:22, Zachary Turner <zturner at google.com> wrote:
>> Is it possible to bisect the revision range?
>
> I've been trying, but discovered that the issue only shows up when the
> full test suite is run (i.e., ninja check-lldb) -- which makes for a
> very slow process. If I run python dotest.py directly the tests do not
> timeout. That's a big clue itself, and I'm starting to suspect it may
> be a problem in FreeBSD that's somehow triggered by a change in the
> range. The hunt continues...

It turns out this is a bug in FreeBSD's timeout(1), which gets stuck
if the child it is directly monitoring does not collect its own zombie
children. Presumably this condition now exists as a side effect of
some change. Even though we may want to investigate and avoid that in
LLDB or the tests, the underlying FreeBSD bug is what really needs to
be fixed.