Description
The examples from the manual add a lot of coverage. It would be nice if we could run some or even all of them on travis (I have no idea how expensiv they are).
I guess the first step for this would be to improve the way we handle them. Right now, there is a relatively convoluted rule in the primary Makefile, which runs some GAP scripts (doc/ref/extractexamples.g
, doc/ref/runexamples.g
, and the same for doc/tut
) and restarts GAP for each chapter. It also runs the tests three times, once starting GAP normally, once with option -A
, and once with a LoadAllPackage()
call added in.
I think we should instead have these scripts in tst; one script which takes one of the manuals (doc/ref, doc/tut, in the future perhaps also doc/dev), and extracts all examples from it into a subdir, say tst/doc/ref
or tst/doc-examples/ref
(and the same for tut). Initially, this would create one .tst file pr chapter, but of course we could revise it.
Afterwards, we can just use TestDirectory
to run the tests. This way, we unify the way the tests are run, and can investigate how long the tests of each chapter take; we could then add one or multiple travis builds which run the examples of different subsets of the chapters (covering as many as possibly, ideally all).
Note: Restarting GAP between chapters might still be advisable, in that case we'd not useTestDirectory
, but the rest of what I said would stay valid, I think.
Activity