[DBD-SQLite] Status report

[Dami Laurent (PJ)]

>-----Message d'origine-----
>De : Kenichi Ishigaki [mailto:kishigaki at gmail.com] 
>Envoyé : vendredi, 24. juillet 2009 05:14
>À : dbd-sqlite at lists.scsys.co.uk
>Objet : [DBD-SQLite] Status report
>
>Hi. As there're several active bug reports and new features,
>I think it may be useful to make it clear what I'm taking or
>pending now.
>

Hi Kenichi,

Thanks for the report.

>1) As Dami's new collation_needed implementation turned out
>to be leaking badly (as noted in a comment, we should free
>the struct somehow), I tentatively disabled it, looking for
>a better implementation (I may also need to disable no_accent
>collation test which is somehow broken now).
>

To avoid leaks, the best thing would be to get rid of the struct
collationNeededInfo, but I'm not sure this is possible. I introduced
it because collation_needed(...) stores only one single pArg, and 
we need two pieces of information (the callback, and the perl dbh).
I tried to see if there would be a way to store the perl dbh in
some private field of the "sqlite3" handle, but it doesn't seem 
to be possible. So if we keep the struct, I think it should be freed
a) at disconnect; b) if there is a second call to collation_needed(...).

I can volunteer to work on this, but don't know how to detect leaks; what are
your tools ?

On the other hand, if no solution is found and collation_needed(..) is dropped,
then a couple of places need to be adjusted (the whole "COLLATION" section
in the doc, and the way builtin collations 'perl' and 'perllocale' are installed).

>4) For the improper utf8 tests issue, I'm wondering what's
>the best solution. Just removing 'require utf8' line is easy
>and I know the tests work fine without that line, but is
>that good enough?
>

As far as I know, this should be good enough. Maybe, for avoiding any
ambiguity, it would be even better to drop the 'utf8::upgrade(..)' calls, 
replacing them by Encode, or by hardcoded literal strings in utf8 (using \x notation).