I learn the weirdest things from Travis CI testing code against many Racket versions, like say from 5.3.5 through 6.8.
For example I just learned (apparently) that backtick is a legal char for an identifier — but only starting with Racket 6.5.
I learned this because I had an at-exp like this: @~a{can't find definition of `@v` in @where}
That instead should be @~a{can't find definition of `@\|v\|` in @where}
to run on older Rackets. ¯_(ツ)_/¯
My biggest build matrix is the one for racket-mode, which is 8 versions of Racket times 3 versions of Emacs: https://travis-ci.org/greghendershott/racket-mode/builds/205669381
It would be really nice to have racket be a real language in travis CI
The build time savings alone would be wonderful
@notjack Well the jobs are < 1 minute overhead to install Racket. e.g. A project with trivial tests: https://travis-ci.org/greghendershott/sha
Whereas in that racket-mode matrix you’re also seeing time to download Emacs source and do a make install
! https://github.com/flycheck/emacs-travis/blob/master/emacs-travis.mk
<1 minute across eight versions is still a decent amount of time. Time savings aside though, it’s more that I don’t like mucking about with the travis script as often as I seem to need to do
Fair enough. Would you like to take the baton? :slightly_smiling_face:
If only I had the time :)
I did look over the guidelines for community-supported travis languages
One thing they mention is needing at least 3 maintainers who will support the integration