[DBD-SQLite] test file 3 failure around todo (was Re: updating bundled SQLite to latest)

[Darren Duncan]

Kenichi Ishigaki wrote:
> Thanks for the heads-up. I updated the bundled library
> to 3.6.11 (and fixed t/08create_function.t issue, too).
> 
> As for how to update, I run "getsqlite.pl" in the trunk
> with <sqlite version to download>, and do some tweaks
> if I find something wrong (like adding new files to the
> repository, updating extensions, etc).
> 
> But I think we should switch to the amalgamated source
> like DBD::SQLite::Amalgamation in the near future.
> It's much easier to update.

Thanks for your quick response.

And I see you joined the list so no direct CC needed.

I agree that we should go like DBD::SQLite::Amalgamation though I understand 
from an IRC chat last night that there are unresolved issues with XS 
compatability, with it not just being a drop-in replacement, so it may take 
attention from someone good at C.  Mind you, Audrey must have gotten it to work 
so in the near future we should see about merging in those changes.

However, unless substituting it seems quite simple, I suggest putting that off 
until our current version is stabilized and released production ready.  That 
way, it would be easier to see empirically if we get various test failures from 
the wider user base, whether it is due to changes for the Amalgamation or 
whether it is unrelated to that.

On another matter, I am getting a test failure on my machines with the CPAN 
release.  Mac OS X 10.5.6 on 2 different arch, both Perl 5.8.9 and 5.10.0 (self 
installed, not system), all with the latest DBI version.

File 03insert.t returns not ok 9 with comments:

# Test 9 got: <UNDEF> (t/03insert.t at line 18)
#   Expected: "4"
#  t/03insert.t line 18 is: skip($unless_min_dbi, $dbh->last_insert_id(undef, 
undef, undef, undef), 4 );

Now this might be a known issue except that so far the CPAN testers say all 
tests passed.  So I'm thinking maybe the tests need a newer version of Test.pm 
or something, but the DBD::SQLite Makefile isn't requiring it.  I have version 
1.25 with my 5.10 and my 5.8.9.  Hmm, maybe that's not it, but then what is 
different about the CPAN testers that reported in, other than none of those 
being Mac OS X?

Is this a known problem with any of you?

-- Darren Duncan