develooper Front page | perl.perl5.porters | Postings from March 2013

Re: [perl #116961] Data::Dumper 2.143 tests failures on 5.10.1

Thread Previous | Thread Next
Peter Rabbitson
March 7, 2013 11:43
Re: [perl #116961] Data::Dumper 2.143 tests failures on 5.10.1
Message ID:
On Thu, Mar 07, 2013 at 03:23:18AM -0800, James E Keenan via RT wrote:
> On Wed Mar 06 21:36:36 2013, wrote:
> > On 03/07/2013 03:55 AM, James E Keenan via RT wrote:
> > > On Sat Mar 02 19:58:34 2013, wrote:
> > >
> > >>
> > >> On older Perl's, XS doesn't load because of a missing symbol:
> > >>
> > >>      Symbol not found: _isWORDCHAR
> > >>
> > >
> > > Does this mean that, going forward from 2.143, we should not support the
> > > XS version when compiled on Perls older than 5.14.0?
> > >
> > >  From a maintenance point-of-view, I would endorse that.
> > 
> > That would cripple the module.
> > 
> Well, do you have a way around the problem David describes above?
> I am in the process of restructuring the test suite so as not to mix
> tests of the XS and pure perl versions inappropriately.  That will
> resolve most of the test failures, but not all of them.

Please do not do this. For one you could inadevertently insert mistakes 
into the tests , and moreover this is the wrong thing to do overall. 
Here is my train of thought:

This is not an ivory-tower nice-to-have - after all there is code in the 
wild (and probably even on CPAN) which uses D::D as a poor man's Sereal 
or some-such. Telling them "you need to upgrade this module, oh and btw 
your serialization code is now 100x slower" is... I can't find the right 
expletive ;)

Also it is frustrating because of the actual barrier to entry for a 
profficient C programmer - it does look like a much more trivial fix in 
C-land, an ifdef or somesuch (but again - my C is laughable). So while I 
or anyone else could spend time "fixing" the tests, it is ultimately 
wasted effort. Mainly because it will take someone with sufficient 
knowledge of the machinery (ideally the person who broke it in the first 
place) about 30 seconds to decide where the ifdef needs to live.

Please reconsider this paper-over, this is not progress by any stretch 
of imagination.

Thread Previous | Thread Next Perl Programming lists via nntp and http.
Comments to Ask Bjørn Hansen at | Group listing | About