[dmd-internals] This is why the auto tester is important

[Iain Buclaw]

On 11 June 2012 10:54, Alex Rønne Petersen <xtzgzorex at gmail.com> wrote:
> Hi folks,
>
> http://d.puremagic.com/test-results/pulls.ghtml
>
> Something like two browser pages of these were 100% green before the
> master breakage. It's extremely hard to keep track of what pull
> requests are passing whenever master breaks.
>
> I'm not writing this just because I want to whine about red builds for
> the sake of whining. I'm writing this for practical reasons: Whenever
> master breaks and all pull requests are reset to a red state, we have
> to wait for the *entire* pull auto testing queue to process before we
> can start gaining an overview of what's currently mergeable and what's
> not. And for the record, this is not a fast process. It takes days,
> literally. Our pull request review/merge/reject process is slow enough
> as it stands, and this is not helping. It's very frustrating, as a
> contributor, to see something this trivial slow down possible merging
> of pull requests. I imagine I am not alone in feeling this way.
>
> All I'm asking is that a look be given to a pull request's state in
> the auto tester *before* it's merged, so that we don't slow everything
> down for everyone unnecessarily. Brad's GreaseMonkey script[1] makes
> this even easier than just opening a separate browser tab. I don't
> think it's an exaggeration to say that it takes less than 10 seconds
> to check a pull request's state.
>
> (I think I ought to mention that I have been the cause of master
> breakage at least two times, if memory serves me right. I didn't
> realize the impact it had until I started reviewing pull requests. Now
> that I actually rely heavily on the pull auto tester, I've realized
> just how annoying these master breakages can be when reviewing pull
> requests (setting aside the issue of master being, well,
> broken/unbuildable; this is easily rectifiable, but getting all the
> pull requests into a green state again is very much *not*).)
>
> So, please, let's all use the auto tester infrastructure properly. It
> should not be a guideline, but a *requirement* to check the auto
> tester before something is merged, IMHO.
>
> Thanks and regards,
> Alex
>

Big +1 from here.

-- 
Iain Buclaw

*(p < e ? p++ : p) = (c & 0x0f) + '0';