CI stuff failing

[Hal Murray]

Matthew.Selsky at twosigma.com said:
> It's an asciidoc implementation written in python 3, instead of ruby.
> We're only testing that the asciidoc program runs without errors. We're
> not checking that it produces the expected output. 

Thanks.

Yup.  Testing documentation is an interesting problem.


> These jobs are only run when the source is "schedule" (eg, the weekly
> scheduled run).  If we think the option testing is valuable, then we can
> run it on every commit.  I'm not sure why we have the "make-tarball" CI
> target.  We don't use that tarball for anything. 

Thanks again.  Is there a way to manually kick them off?  Or get them to 
run an hour to 3 after a commit?

There seems to be a pattern of running standard and with-refclocks on 
various platforms.  What's the overhead of of a test?  Would it be 
appropriate to make a mini option tester to be run on each platform?


-- 
These are my opinions.  I hate spam.