[CDBI] Re: Transaction over parent and children has_a objects

Peter Speltz wrote:
>> Same thing. Figured it out though. When i removed an Ima::DBI patch i got
>> from you in email it worked.

Perrin Harkins replied:
> That's strange, because the patch is really pretty simple. It works
> the same as before, unless you change process IDs between calls to
> db_Main(). It has been working well on a moderately large site where
> we use transactions in the same way you're showing.

I agree. I don't see anything in Perrin's patch that should cause this
problem.

Could it be a MySQL thing? I know there have been some DBDs (notably older
versions of DBD::Pg, I believe) where ping() doesn't give the correct
response under some circumstances and a new dbh is created, which breaks the
transactional model.

Nesting transactions could also give the same error, but Peter said it worked
when he reverted his Ima::DBI, so that's not the problem obviously.

Later,
Ed