SciRuby / nmatrix

Dense and sparse linear algebra library for Ruby via SciRuby
Other
469 stars 133 forks source link

Add Ruby 2.5 and 2.6 to travis configuration #619

Open translunar opened 6 years ago

translunar commented 6 years ago

From #617 it came to our attention that our Travis configuration is a bit out-of-date.

610 is slightly different — this is correcting the existing configuration, which still needs to be done. But this issue, here, is just about adding 2.5 and 2.6 to the existing broken configuration.

ngetahun commented 6 years ago

Hey @mohawkjohn, is this issue still up for grabs? I'd like to make a PR.

translunar commented 6 years ago

Yes, please! The Travis configuration is a mess. Take a swing!

translunar commented 6 years ago

@ngetahun Did you end up having any luck?

leesharma commented 5 years ago

Hey there! It looks like this issue has been sitting for a few months—mind if I give it a go?

translunar commented 5 years ago

By all means!

On Mon, Oct 1, 2018 at 2:11 PM Lee Sharma notifications@github.com wrote:

Hey there! It looks like this issue has been sitting for a few months—mind if I give it a go?

— You are receiving this because you were mentioned. Reply to this email directly, view it on GitHub https://github.com/SciRuby/nmatrix/issues/619#issuecomment-426064409, or mute the thread https://github.com/notifications/unsubscribe-auth/AAKFpKnLevbEq1tZDZmFtD2lO7D0Sw5oks5ugoUIgaJpZM4TMqRI .

leesharma commented 5 years ago

It looks like there's a compatibility issue between packable and ruby versions ≥ 2.5 (see SciRuby/packable#4). Because of this, tests that use their monkey-patched methods fail in CI for 2.5.1 and 2.6.0-preview2 (an example failure with the same error).

From the discussion on that issue, it looks like the problem isn't going to be resolved anytime soon. Luckily, the tests pass locally for me on OS X and Ubuntu; it looks like the problem is limited to CI and Debian systems. I think your options are:

How do you want to handle this?

translunar commented 5 years ago

I'd say mark as pending on affected systems. It's difficult to extricate packable, but it's really only used in the Matlab IO, which should eventually become a plugin.