Fix Travis problems (2019 Oct)

[wibeasley]

It looks like Travis has not passed in two years (ie, Oct 2017 was the last full success).

Things pass on my local machine (which is Ubuntu, like Travis), so I think the problem likely lies with the travis config, rather than the R machinery. I'll try to

• [x] simplify the config until it passes, then
• [x] re-add the Travis matrix cells/conditions so it passes, but still replicates the intent of the current config

[wibeasley]

Does anyone strongly believe the 7.3 OSX build should continue? What about Ubuntu 12.04 LTE (ie, 'precise')?

I think these older releases could be the cause of the travis fails. The 7.3 tool chain is no longer checked on CRAN (only 8.2). Similarly, Ubuntu 12.04 was released 7.5 years ago, and was EOLed 2.5 years ago.

[pdurbin]

@wibeasley I'd say you should kill those builds for older operating systems if they're causing you trouble. If people complain, we can always attempt to add them back in later.

[wibeasley]

@pdurbin sounds good. The config now has the two most recent Ubuntu versions, plus the 8.2 toolchain on OSX. And I agree that if people express the need to check additional versions, we can easily evaluate the tradeoffs of resurrecting those checks.

I think this is what you meant, but this controls just the checking on the master branch. CRAN handles the checks and binary builds on CRAN.