appveyor: skip older pr builds

[ghost]

No description provided.

[ghost]

request: trivial

[pradyunsg]

@xoviat Could you trigger 2 builds on this PR such that one would get cancelled due to these changes?

[ghost]

@pradyunsg Done.

[pradyunsg]

@dstufft @pfmoore Could one of you take a look at this?

[pfmoore]

It looks like it does what it says (in other words, the Powershell command added appears to do what the comment suggests it does) but as to whether the idea is good, I don't know.

The Appveyor tests are only the unit tests, so they typically don't take anywhere near as long as the Travis tests (even after recent improvements). So I'm not sure the time saved here is that important. There's 8 appveyor tests, each taking just over a minute, And if I understand the point of the change, it's only likely to make a difference if within that 8 minute window, the author pushes an extra commit to the PR. That seems fairly unlikely, if I'm honest.

I've no objection to improvements like this in principle. But given the shortage of Windows (and specifically Powershell) expertise among the pip developers, I'd be inclined to keep the Appveyor scripts as simple as possible, and hence reject this PR as unnecessary.

[pradyunsg]

@pfmoore

There's 8 appveyor tests, each taking just over a minute

I am not sure if you saw the AppVeyor builds for #4589? Those held up the build queue for what was a few hours since it caused a regression in pip resulting in a hung tox.

I believe that it was the motivation for making this change - which I too agree to; other PRs did not have their builds run until that build finished, which affected me too.

I'd be inclined to keep the Appveyor scripts as simple as possible, and hence reject this PR as unnecessary.

Sounds fine by me; I don't think we'll get such PRs (that makes changes that hang the build as a result of them) often anyway. That said, if what I said above does change your position and this gets merged, I'm good too.

---

In short, I'm fine regardless of whether this is merged or not, slightly preferring that this gets merged.

[pfmoore]

I am not sure if you saw the AppVeyor builds for #4589?

No I hadn't. If this PR is motivated by that situation, rather than a simple "let's make the tests faster" motivaton, then I see your point and agree this would be good to have.

I'm still a little concerned about the maintenance - let's see what @dstufft thinks, and if he's OK with it then we can merge it.

[ghost]

I actually picked this up from obspy; put in by @barsch in obspy/obspy@fedff4b.

I actually just gave this some thought and I think it might be better to actually timeout on AppVeyor if tests run for more than 5 minutes. Making a PR for that.