I am going to toss out to this list an idea that occurred to me today,
the final day of YAPC::NA::2015 in Salt Lake City.
The concept of the CPAN river emerged at this year's QA Hackathon in
Berlin and was referenced in at least one presentation here at YAPC. (I
believe it was Patrick Michaud who described perl itself as the snows at
the top of the mountain top.) So distro authors/maintainers have to be
aware of the downstream impact of code changes.
In the course of preparing my own talk, I had occasion to check out
List-Compare on metacpan.org and noted that that site, along with
cpantesters.org, has begun to present lists of downstream distros for
which List-Compare is an upstream prerequisite. List-Compare came up
for review in the CPAN Pull Request Challenge this month, and I asked
PRC participant Paulo Custodio to make sure that there was one unit test
in the test suite for each example cited in the documentation. Today it
occurred to me that I ought to survey each of List-Compare's downstream
dependencies and see how each of *them* was using List-Compare and make
sure I had a unit test in my suite for each way the downstream distro
was using L-C.
With the metacpan API, this ought to be automatable. Then, judicious
use of source filters, grep and ack ought to be able to provide a distro
author/maintainer with at least a *first draft* of a t/*.t file which
would represent the current state of downstream uses for that distro.
Is there any prior art for it?
This task is non-trivial -- but how big would it actually be?
Would it be maintainable?
Discuss.
Thank you very much.
Jim Keenan
