[DBD-SQLite] DBD::SQLite and CPAN

[Curtis Jewell]

On Fri, 19 Jun 2009 18:45 +0100, "Ash Berlin" <ash_cpan at firemirror.com>
wrote:
> 
> On 19 Jun 2009, at 18:37, David Dooling wrote:
> 
> > I am trying to clean up the prerequisites for a package that requires
> > both DBI and DBD::SQLite (UR).  CPAN wants to order the prerequisites
> > in alphabetical order, so it tries to install DBD::SQLite first.  It
> > comes to this part of the DBD::SQLite Makefile.PL code
> >
> >  # Because DBI generates a postamble at configure-time, we need
> >  # the required version of DBI very early.
> >  my $DBI_required = 1.57;
> >  eval {
> >          require DBI;
> >  };
> >  if ( $@ or DBI->VERSION < $DBI_required ) {
> >          print "DBI 1.57 is required to configure this module,  
> > please install it or upgrade your CPAN/CPANPLUS shell\n";
> >          exit(0);
> >  }
> >
> > and fails (although the exit status is zero) because an older version
> > of DBI is installed.  Even though DBI later appears and as PREREQ_PM
> > in the WriteMakefile call, that line never gets executed, the Makefile
> > never gets created, and CPAN will go no further with the DBD::SQLite
> > install. CPAN does continue to try to install other modules,
> > ultimately failing the requested install because DBD::SQLite is
> > missing (or too old).
> >
> > I am not sure what the issue around the DBI postamble is so I am not
> > sure how to fix this so DBD::SQLite to play nicely with CPAN.  Any
> > ideas/guidance on how to get this to behave as one might expect?

The DBI postamble copies two files from the DBI distribution into the
DBD driver's distribution - they're the standard framework for any XS
dbd driver.

So the postamble is vital.

I would assume that there was an incompatible change or a necessary bug
fix in one or the other of these two files in the 1.56 -> 1.57 era that
would make compilation impossible, although I could be wrong. (It's
quite possible:
http://search.cpan.org/diff?from=DBI-1.56&to=DBI-1.57#Driver.xst - The
postamble itself hasn't changed since 1.56, so it's not that.)

> Two possible solutions I can think of:
> 
> 1) configure_requires. Doesn't help with old clients

No, but it's the easiest way to help new ones. It can't hurt.

> 2) Use the 'Are we running under a CPAN shell' from the (now improved)  
> Module::Install::AutoInstall.

Do this, too. It will catch what the previous one does not.

> Oh, a bonus 3rd option.
> 
> 3) Remove the check and delay it to the first test which does BAIL_OUT  
> if DBI is too old.

Does it even COMPILE if DBI is too old?

--Curtis
--
Curtis Jewell
swordsman at csjewell.fastmail.us

%DCL-E-MEM-BAD, bad memory
-VMS-F-PDGERS, pudding between the ears

[I use PC-Alpine, which deliberately does not display colors and pictures in HTML mail]