[Dbix-class] ANN - SQLite 3.6.8 adds nested transactions

[Darren Duncan]

Toby Corkindale wrote:
> The disconnection code is borked too; if you connect, do some work with 
> statement handles, and disconnect again, and then repeat many times in 
> the same process, you'll crash out eventually.
> 
> The annoying thing is that there are trivial patches on rt.cpan.org to 
> fix it, and they've been there for ages, and they aren't getting 
> applied. :(

Well, if that's the case, and one or more people can supply compatible patches 
for various problems, then maybe someone with less C-fu than otherwise can then 
co-maintain DBD::SQLite and apply the patches.

For example, it would be easier for me to become co-maintainer if such 
help-me-fix-it feedback is readily available.  And I do want to see SQLite 
succeed with Perl, now and indefinitely, so maybe I'll go ahead and ask for 
co-maintainership.  Any objections or support?  Note that it is likely to be a 
few weeks before I release anything.

In that vein, here's a question:

To your knowledge, were the problems with 1.13+ due to changes in the Perl 
binding itself, or due to a lack of changes in the Perl binding that were 
required to maintain SQLite compatibility?

If the problem was due to changes rather than a lack of changes, I could then 
just use version 1.12 as a point of departure for a new 1.15_01 or what-have-you 
release.

Do any reported problems with 1.13+ also apply to 1.12?

> Do you know if Audrey's fork has fixes applied, or is it just simply an 
> upgraded-to-version-3.6 thing?

I don't know about fixes; haven't looked close enough at it.  But I do know that 
the source files are significantly different.  The main point of the 
amalgamation source, which is supplied officially by SQLite, is to let less 
capable dev environments like Windows make and build it more easily, because 
lots of pre-compilation work has been done already.

-- Darren Duncan