David Cantrell wrote:
On Tue, Nov 06, 2007 at 05:20:15PM +0000, Martin J. Evans wrote:
It appears from "Can't locate DBI.pm" that DBI is not installed but the
Makefile.PL contains:
$opts{PREREQ_PM} = { DBI => 1.21 }
Your Makefile.PL doesn't get that far. It gets as far as the "use DBI"
on line 15. You can't make any assumptions about what is installed
until *after* Makefile.PL has run.
Thanks, I missed that - I only took over DBD::ODBC recently. The actual
problem is you need DBI because you need a function from DBI to locate
DBIXS.h etc in order to create the Makefile.
so long as ExtUtils::MakeMaker version is >= 5.43. If I put
$opts{PREREQ_FATAL} = 1 in the Makefile.PL will this stop this error
being reported again.
You can put "i am a herring" in there for all the difference it will
make :-) Your Makefile.PL will still break at line 15. The solution
is rather involved:
eval "use DBI 1.21";
if($@) {
... stuff that uses CPAN.pm to install DBI goes here ...
}
eval "use DBI 1.21";
if($@) {
warn("
Something's badly broken with your environment, I couldn't
install DBI
");
exit(0);
}
hmm. Are you saying if I did this properly (as above) then the cpan
testers tests would get past this? Do you know of another module which
does this as above? As far as I can see this is the first time DBD::ODBC
has failed in this way from cpan testers and yet it has been this way
for years. I greatly appreciate the cpan testers work (this time it
picked up a bug with using older perls for instance) so I'd be happy to
make any changes to make cpan testing better. Perhaps I need to do some
research on cpan testing to see what is recommended.
Do that for every non-core module you want to use in Makefile.PL.
Thanks for all your efforts.
Martin
