CI:ECL: `load` with `:verbose` and `:print` options writes text of file unsafely about common-lisp HOT 5 CLOSED

That sounds like an interesting implementation and I'd love to see it. (It does sound like it might go against my "don't be clever in example.lisp" pseudo-rule though...)

If setting *print-circle* to nil fixes it then do it in xlisp-test - perhaps under a feature test for ECL.

from common-lisp.

Let me know if you think it's too clever.

wobh@1982134

I also added a bunch of tests.

from common-lisp.

It's definitely ECL's load with the :print option and setting *print-circle* to t makes it work as expected. Reviewing the spec for *print-circle* http://l1sp.org/cl/*print-circle* it seems like it would be fine to set it to t generally, and perhaps even a good idea.

from common-lisp.

A passing build: https://travis-ci.org/wobh/xlisp/jobs/83063702

I have two ideas here based on the same foundation.

First add some constants to label the verbosity levels. This takes care of an irksome remainder I decided not to do initially for some reason, but also makes distinguishing the two solutions a easier.

Solution 1. Set *print-circle* to t in the lexical scope of load-package. This is a minimal typing solution and the passing build above is the prime-factors exercise rebased onto that branch. At the bottom you can see the circular list printed correctly. The rationale here is that whatever the load outputs, it should be safe, so just set it to t and be done.

Solution 2. Set *print-circle* to the outcome of (verbosity-p +debug+) in the lexical scope of load-package and set the load :print argument to the same. This is a minimal effect solution, observing that the extra output of load with :print is not strictly desirable for general testing, and *print-circle* doesn't need to be any value not required.

I prefer solution 1 for the time being, but I think considering what solution 2 does for us is worth thinking about more generally. If/when we find another place to set *print-circle* to t or have to set it's dynamic value in test-exercises we can do so by verbosity level required with verbosity-p.

from common-lisp.

I agree - it looks like setting *print-circle* to t is a good idea. It is a good thing for a testing harness to do - given that there may be circular data - we'd want the harness to be able to handle and report on that without it having its own problems.

I go with solution 1. If we tied it to verbosity then we should consider the other *print-x* settings which might also be needed/useful. But I think the simpler approach is better for now.

from common-lisp.