[DBD-SQLite] SQLite 3.7.x problem fixed, please dev-release

[Toby Corkindale]

On 26/08/10 12:35, Darren Duncan wrote:
> Toby Corkindale wrote:
>> On 26/08/10 05:38, Darren Duncan wrote:
>>> Under the circumstances, I recommend releasing what we have now as
>>> version 1.31 production within a week if there are no showstopper
>>> problems discovered with 1.30_04 in the meantime.
>>
>> It sounds like there are some quite major changes in this version of
>> SQLite, and thus also DBD::SQLite.
>>
>> As with other major releases, I'm interested in testing the dev
>> release against our code which does tend to stress SQLite - however
>> this isn't something I can fit in in the time frame you're discussing
>> ("within a week").
>>
>> I am sure I'm not the only person in this situation.
>>
>> Do you really need to release a so-called "stable" version so rapidly?
>> If what you have now is a release candidate, then would you mind
>> leaving it for a bit longer while people have a chance to shake it out?
>>
>> DBD::SQLite is a major piece of many applications. Please be careful
>> what you nominate as "stable".
>>
>> Thanks for your efforts in keeping it up to date and incorporating
>> upstream improvements - it's really appreciated.
>
> Keep in mind, first of all, that most bugs of any consequence would be
> in SQLite itself. SQLite itself has already had 3 patch releases since
> 3.7.0 which introduced the major changes, and 3.7.0 came out a full
> month ago. Meanwhile 3.7.0.1 came which fixed a few bugs, then 3.7.1
> with a greater number of changes, and then 3.7.2 which just fixed a bug.
> So the codebase with the most serious likelihood of issues has already
> had a month of shake-out, and moreover the SQLite itself has 100% test
> coverage.

I understand that most of the issues would be in SQLite rather than 
DBD::SQLite, but the Perl world hasn't tested that either yet.

I appreciate that both have unit test coverage, but just testing each 
line of code doesn't guarantee you find odd edge cases in real-world 
situations, nor odd interactions on particular operating systems, perl 
or compiler versions.

> Granted, DBD::SQLite has relatively little test coverage, and mostly
> piggybacks on assuming SQLite itself or Perl itself is fine.
>
> That all being said, how much time do you need to do your stress test?
> Two weeks?

I and my colleagues should be able to get some testing done in the next 
fortnight, sure.

I suspect many people will only catch the announcement via your posts on 
other DB-related lists, so may take a while to get around to testing as 
well.

Things will probably be OK in the end, but I do want to suggest that 
there isn't a huge rush to get a "stable" version released to CPAN, so 
maybe when there are "major" released like this, you could give people 
weeks rather than days to test them in?

Thanks for your understanding,
Toby