nqp-rx tests: in 'make test' or in 'make fulltest'

[Reini Urban]

On Wed, Dec 5, 2012 at 9:34 PM, James E Keenan <jkeen at verizon.net> wrote:
> Today, in 53cf923e2, tests in ext/nqp-rx/t/nqp/ were added to the Parrot
> distribution.
>
> Do we need to have these in the 'make test' target -- which means they will
> be run each and every time someone tests Parrot?
>
> Or would it suffice to include them in the 'make fulltest' target?
>
> I'm leaning to the latter, in part because I really don't want to add to the
> running time of 'test', and in part because I don't know whether they will
> be significant enough indicators of breakage to warrant being run each and
> every time.
>
> What do others think?  (Dare I suggest we convene a parrotsketch for this?
> ;-) )

I added them to the compiler testset, because:
1) tests from other compilers are also in the compilers testset,
2) parrot-nqp (aka nqp-rx) shares the testsuite with nqp (the new generation),
3) the tests were there previously already, but were never executed.
So nobody caught the bitrot,
and nobody merged the new tests from upstream. This is unacceptable.
rakudo is our prime customer
and we should be happy to be able to execute some of their tests easily.
4) the tests are very short

To the problem - GH #874:
I've only learned recently that nqp and nqp-rx create uncatchable
compile-time errors when icu is not present and the src code contains
\cxx charnames. There's no nqp BEGIN block, only a INIT block.
No way to skip the test inside the test.
So we must skip one test from the suite if parrot has no icu, in all
three nqp repos.
parrot/ext/nqp-rx, perl6/nqp and perl6/nqp-rx

-- 
Reini Urban
http://cpanel.net/   http://www.perl-compiler.org/