Dropping numpy 1.7 and 1.8 testing - #6003

In the past, @MSeifert04 had been pretty good on removing stuff like this... ping! 😂

In the past, @MSeifert04 had been pretty good on removing stuff like this...

Sounds like I'm some sort of vacuum cleaner. Do you mean for the workarounds or the travis matrix? 😄

At a casual glance:It seems we're not testing numpy 1.10 (at all) and python 3.5 (except on OSX). Is that correct or is that implicit in one/several of the other tests?

I meant "removal of np 1.7 and 1.8 workaround". 😄

@MSeifert04 - 1.10 is tested in the image tests, and indeed 3.5 is tested with OSX. I think this should be sufficient enough for now, as we probably will drop 3.3 for the v2.0 release anyway (though I didn't want to run ahead and do that in this PR until the APE change is accepted).

Re code workaround removal: I can give that a try, too if you don't have time, but I didn't aim to do that in this PR, as this should be quick one to get in asap to let things go ahead in #5819.

@bsipocz - I'm happy to make a PR to your PR in the next 20 minutes that removes numpy <=1.8 cruft. Let me know if you are already doing it, though.

@mhvk - Please go ahead, I haven't started doing it. Also, you should be able to push directly to this branch, no need to open a PR to the PR.

I'm done, but don't actually know how to push to your branch from the command line...

p.s. The drop of numpy 1.7 in particular is nice; also, we're left with just three real work-arounds (a handful if one includes tests).

@mhvk - you need to have my fork as a remote (assuming you call it bsipocz):

git push bsipocz your_branch:travis_drop_np_17_18

or if you are on the top of the branch you made the clean:

git push bsipocz HEAD:travis_drop_np_17_18

I pushed to your branch as you suggested, but it doesn't seem to appear here...

I thought it might be perhaps easier if I just took control... see #6006. Will close this one...

@mhvk - This is weird. Which protocol do you use for remotes? I switched to use the same as for my clone, git@github..., as some others didn't work.

I did use the ssh one, and I think on your branch travis got started, but it didn't seem to propagate to astropy; possibly, I was just too impatient and should have waited for the tests to pass, so that then the PR would be propagated?

Oh, now I see. Somehow you pushed the branch to astropy (upstream) rather that to my fork and travis started to run on that: https://travis-ci.org/astropy/astropy/branches

Do you still have the command you used? I would be interested in getting this sorted out either here or on gitter, as pushing to other's branches is quite a powerful tool and would be a waste not so share the howto with you.

Good that you asked about what I actually did, as it is now quite clear why this failed:

git remote add bsipocz [email protected]:astropy/astropy.git
git push bsipocz HEAD:travis_drop_np_17_18

Here, I think I just looked at your page and clicked on the astropy repository; on my page, it is set up to bring me to my own fork, but on yours I now see it simply links to astropy/astropy.

So, clearly just me not looking carefully at what I was doing...

Glad to see that this wasn't an obscure bug somewhere, and I'm pretty sure you will like using this feature.

I've now deleted the branch from astropy/astropy.