Front page | perl.perl5.porters |
Postings from May 2007
Re: Perl in LSB 3.2
From: Stew Benedict
May 17, 2007 05:32
Re: Perl in LSB 3.2
Message ID: Pine.LNX.firstname.lastname@example.org
On Thu, 17 May 2007, Andy Armstrong wrote:
> On 17 May 2007, at 12:07, Stew Benedict wrote:
> > I don't own the book either. While it may indeed be a good language
> > reference, for LSB we would need something a little more accessible,
> > online, as well as something more along the lines of a formal
> > specification document (In the context of exactly what we decide to
> > specify).
> Historically Perl has always been specified as (roughly) 'whatever
> the Perl interpreter interprets'. That probably doesn't help you much :)
> I suspect that fact is crucial to the personality of Perl. It's
> certainly an accepted part of the culture and gives rise to memes
> such as "only perl can parse Perl".
> So there's no notion of a language specification that's independent
> from the current implementation.
> As Nick says the closest thing we have is probably the test suite. Do
> you think it's adequate to get the test suite running against the
> installed Perl and then have the standard say just 'must successfully
> execute the test suite that can be found at XXX' ?
It could be. LSB certainly doesn't want to specify more than is
We could simply specify the interpreter to be present on the distribution,
and that it passes the test suite. That's toward one end of the spectrum
of possibilities, and if that's all that makes sense for Perl, we can
leave it at that.
Typically, we would package and provide the test suite as part of our
"runtime test bundle", as well as putting individual tests into rpm/deb
repos for download. If the tests are from an upstream source, we set
our builds to pull from that source and use it, applying any packaging
tweaks needed to install it in an LSB conforming place (/opt/lsb/test/foo)
and to have it work with our testing setup: