On Sun, Sep 13, 2009 at 12:26:57PM +0200, H.Merijn Brand wrote:
> On Sun, 13 Sep 2009 11:12:38 +0100, Tim Bunce <tim.bu...@pobox.com>
> > > - if (!filename || strEQ(filename,"STDERR")) {
> > > + if (!filename || strEQ(filename,"STDERR")
> > > + || strEQ(filename,"*main::STDERR")) {
> >
> > Wouldn't it be better to detect when a glob is being passed and use the
> > filehandle in the glob? That way it would work for any *FOO fh.
>
> That's what Nick said too. I am looking onto this, but as STDERR was
> being referred to here specifically, I thought this at least fixes the
> special case for all backwards compatibility issues. (My tests suddenly
> started creating files with the name "*main::STDERR", and upgrading DBI
> is way easier than upgrading perl.
I'd be as happy for names starting with * to trigger a warning.
> Feel free to beat me in doing it right though.
That's not going to happen :) NYTProf v3 and java2perl6 have a higher
priority call on my limited free time at the momement, along with a
bunch of conferences in the next few weeks, so I'm especially grateful
for any help with DBI.
> Is it OK with you if I upgrade DBI to be using the Devel::PPPort-3.19,
> so we're more compatible with more recent perl's?
Yes, please do.
> This is quite a big change if I follow all the advices
>
> File needs grok_number, adding static request
> File needs grok_numeric_radix, adding static request
> Files [./DBI.xs ./Driver_xst.h] need newRV_noinc, adding global request
> Files [./DBI.xs ./DBIXS.h] need sv_2pv_flags, adding global request
>
> just one chunk:
> @@ -1227,12 +1232,12 @@
> - dbih_imp_rv = newRV(dbih_imp_sv); /* just needed for sv_bless */
> + dbih_imp_rv = newRV_inc(dbih_imp_sv); /* just needed for sv_bless */
> - DBIc_IMP_DATA(imp) = (imp_datasv) ? newSVsv(imp_datasv) : &sv_undef;
> + DBIc_IMP_DATA(imp) = (imp_datasv) ? newSVsv(imp_datasv) : &PL_sv_undef;
That's fine.
Driver authors should do the same, of course, but I don't think they'll
be any problems if DBI does but they don't.
> $ git status
>
> also shows me that t/zv* is not tracked. Is that correct in svn too?
Yes. They get deleted by make clean.
Thanks!
Tim.
p.s. Don't forget to credit yourself in the Changes file!
