On Jul 14, 2006, at 7:34 PM, brian d foy wrote:
>> Anyone have a recipe for building DBD::ODBC?
>
> I built it yesterday on Solaris 10 without incident. What's the
> problem
> you're running into?
I've got it building with -o /usr. There are a lot of "signed vs.
unsigned pointer mismatch" type warnings - but those are new warnings
in GCC 4.x, and I've seen them in a ton of stuff. I don't know if
they're relevant.
This may be a clue:
"Umm, this looks like a unixodbc type of driver manager."
It's not - I'm configuring against the built-in iodbc in /usr, not
unixodbc. And the tests fail with a dynamic linker "symbol not
resolved in dynamic_lookup" error, which could easily be the result
of using the wrong API.
What I'm aiming for is a binary build I can include with the next
CamelBones patch release. Frankly, I don't know a whole lot about
ODBC on *nix - which is a big part of the problem, I'm sure. I need
to hit the books.
sherm--
Web Hosting by West Virginians, for West Virginians: http://wv-www.net
Cocoa programming in Perl: http://camelbones.sourceforge.net
Sherm Pendley - 15 Jul 2006 01:34 GMT
> I've got it building with -o /usr. There are a lot of "signed vs.
> unsigned pointer mismatch" type warnings - but those are new
[quoted text clipped - 9 lines]
> resolved in dynamic_lookup" error, which could easily be the result
> of using the wrong API.
Aha! Yes - that was a clue.
There's a /usr/lib/libodbc.a that's a symbolic link to /usr/lib/
libiodbc.a. As it happens, DBD::ODBC uses the presence of /usr/lib/
libodbc.a as a sign of unixodbc, and moving the symlink out of harm's
way allowed Makefile.PL to correctly detect iodbc instead.
The 01base.t test is working now without the unresolved symbol
errors. Now I'll have to set up an actual DSN and run a real test
against it.
sherm--
Web Hosting by West Virginians, for West Virginians: http://wv-www.net
Cocoa programming in Perl: http://camelbones.sourceforge.net