Re: texi2html tests are huge

[Patrice Dumas]

On Sun, Jan 15, 2012 at 03:40:20PM -0800, Karl Berry wrote:
> 
> Also, as I said before, it is ok with me to drop the whole idea and just
> run all the tests every time.  Since, as we already know, users will
> generally not have the prerequisite packages and therefore no tests will
> be run at all.

I would not prefer that.  The normal tests already take 4 minutes to run 
on my computer, and I guess that the test suite run time can only increase 
and never decrease (except if makeinfo is sped up, of course...).  The 
long tests also takes 4 minutes to run, so the gain in development time
of not having all tests run each time is big.

>     many corner cases are tested that should not happen in real life manuals.
> 
> If a division is made, according to my original theory I think it would
> be better to divide them into "tests that take a long time to run" and
> "tests that are quick", rather than "tests that are important"
> vs. unimportant.

But that is not really a meaningful division.  In the main test suite, that
takes 3 minutes and 30 sec to run there is no long test, but more than 
860 (each one with 3-8 things tested) for the tests using t/test_utils.pl 
(the other tests are negligible).  So the time to run is not a useful 
division.  The tests in tp/tests that test the texi2any command are also 
numerous but not onerous, since the 49 tests take about 40 seconds.
Moreover, in the main test suite, tests are organized along texinfo command
groups or converter type.

-- 
Pat