Re: [Boost-testing] [EXTERNAL] Develop regression shunted, failing process_jam_log compile, FreeBSD

Subject: Re: [Boost-testing] [EXTERNAL] Develop regression shunted, failing process_jam_log compile, FreeBSD
From: Belcourt, Kenneth (kbelco_at_[hidden])
Date: 2014-09-18 23:36:43

On Sep 18, 2014, at 9:08 PM, Ben Pope <benpope81_at_[hidden]> wrote:

> On Friday, September 19, 2014 01:30 AM, Belcourt, Kenneth wrote:
>> 
>> On Sep 18, 2014, at 11:14 AM, Ben Pope <benpope81_at_[hidden]> wrote:
>> 
>>> On Friday, September 19, 2014 01:07 AM, Belcourt, Kenneth wrote:
>>>> 
>>>> Can you tell me if you remove the boost_root/boost directory before you run your tests?  My Linux testers remove boost_root entirely before running but they still run okay.  Our Darwin testers also remove boost_root but they all fail.  This seems to indicate an OS / toolset kind of problem so that’s why I ask.
>>> 
>>> I just removed that directory on BP x86_64 C++11 (Ubuntu), let's see how it goes.  It'll build after master, which is building now.
>> 
>> Thanks Ben, that will be a useful data point.
> 
> Looks like it failed in a similar way:

This convinces me that we need to revert this MPL commit and figure out what is actually broken.  I know Andrey’s posted a PR to “fix” this but I’m really concerned we still have no ability to clearly articulate why the various testers / developers are broken.  For example, some Ubuntu testers are broken, others are working.  Same for RHEL 6 users (I’m okay but PNNL broke).  No apparent rhyme or reason so that scares the heck out of me.  At least Darwin is completely broken for everyone, and Windows appears to be completely broken along with FreeBSD and QNX.  I’m stumped.

Thanks for the info.

— Noel