Subject: Re: [boost] [winrt support] Adding support for Windows 8 store/phone to Boost libraries
From: Edward Diener (eldiener_at_[hidden])
Date: 2014-07-22 10:16:39
On 7/22/2014 6:22 AM, Niall Douglas wrote:
> On 21 Jul 2014 at 22:52, Edward Diener wrote:
>
>>> Steve's pull request for Boost.Thread is being held back by me until
>>> the 1.57 release as I currently don't have the CI facilities to
>>> verify for myself it works to my satisfaction.
>>
>> Boost could really do with the idea that we should be able to run and
>> publish a regression test for some library that is not either the
>> 'develop' or 'master' branch. That way someone with facilities to test a
>> change could publish regression tests based on that change for others,
>> like yourself in this case, who may want to accept a pull request but
>> naturally needs to know that it tests OK before doing so. While pull
>> requests are a good idea, having to blindly accept one at any time
>> without verifying it for oneself is not as I am sure everyone knows. So
>> I fully understand your problem, but we need a more generalized solution
>> to it.
>
> Both Jenkins and Travis are very easily configured to automatically
> test pull requests and report back onto Github whether they are good
> or not.
Sounds like a firm of lawyers, "Jenkins and Travis" <g>. I will look at
them and see what they offer. Thanks !
>
>> Maybe someone on the Boost Build/Regression Tests side of things would
>> know how to setup and publish a regression test for a single library off
>> of some non-'develop' and non-'master' branch so that the test results
>> are accessible to others.
>
> Jenkins also publishes Boost.Build results just fine. Jenkins has a
> RESTful API which lets some central server interrogate it for results
> for some set of projects. As does Travis, incidentally.
>
> You probably see where I'm going with this.
Maybe this can be integrated with Boost regression testing in some way
that provides results for a branch that is not 'develop' or 'master',
and can be used by anyone with a pull request.
>
>>> Hell, I don't even
>>> have AFIO being CI tested right now, modularisation foobared
>>> everything, but the CI orchestration work being done for AFIO will be
>>> reusable for Boost.Thread. And my Windows CI is now Win8.1 instead of
>>> XP, so now I can bring VS2012 and VS2013 into the soak test runs.
>>
>> I am not even on Win 8 yet, but luckily the latest Visual Studio
>> releases still run under Windows 7. But of course I can't test new
>> Windows 8 features as they relate to Boost until I install Win 8 ( or
>> Win 8.1 ).
>
> I *believe* that the WinRT SDK needs Win8 minimum. It certainly did
> when I worked at BlackBerry, I got to spend a week at home using my
> own computer because BlackBerry didn't have a single Windows 8
> machine to hand at that time :)
That is why I am mildly concerned about changes to Boost for Windows
Store and Phone apps. They need to be tested like anything else to see
if they work properly.