[Kde-perl] PerlQt-3.004-RC2

[Germain Garand]

Le Jeudi 12 D=E9cembre 2002 09:16, St=E9phane Payrard a =E9crit :
> On (11/12/02 21:24), Germain Garand wrote:
> > From: Germain Garand <germain@ebooksfrance.com>
> > To: kde-perl@mail.kde.org
> > Subject: Re: [Kde-perl] PerlQt-3.004-RC2
> > Reply-To: kde-perl@mail.kde.org
> > Date: Wed, 11 Dec 2002 21:24:40 +0000
> >
> > Le Mercredi 11 D=E9cembre 2002 20:29, St=E9phane Payrard a =E9crit :
> > > Hi,
> > >
> > > Still having the link problem with 3.004-RC2
> >
> > It looks like you have a custom/non-standard Qt built (I'd say:
> > QT_NO_MIME) ? In such case (see "Troubleshooting and Configure Option=
s"
> > in the doc ) :
> >
> > perl Makefile.PL --with-qt-dir=3D/home/stef/qt-x11-free-3.1.0
> > --with-treshold=3D0
> >
> > is ought to take care of every vicious #define you may have...
> >
> > G.
> > P.S: did Makefile.PL cured your automake 1.7 problem ? It would be go=
od
> > to know
>
> Hi Germain
>
> With RC1, you suggestion of using aclocal solved the problem.
> With RC2, the Makefile.PL did work.
> In both case, I had the problem at link time.
>
> I think this is a good move to work at making the installation solid.
>

OK, but you did not tell me if your Qt is built with default options or n=
ot.=20
If not, then --with-treshold=3D0 must be used.

The reason for that is simple: Qt as a great flexibility at compile time =
, but=20
as also a big weakness in that respect : it does not "remember" all symbo=
ls=20
that were #defined.

Thus, the only way to know if a given feature is there afterward is to te=
st=20
all dubious symbols, one by one. That's what --with-treshold do : raise t=
he=20
number of testing from a few 10 to more than 130 (I've probably missed a =
few=20
ones but not much).

Now, if indeed having a rock-solid-by-default build system in all situati=
ons=20
is what we need then I can lower the default treshold to zero... that wil=
l=20
add  five to ten minutes to the total build time for Smoke.

mmh.. yeah, I think we'll do that... I've just finished building the PyQt=
=20
bindings (four hours !) and they don't test the #defines at all... so mis=
sing=20
symbols, frustration, and double compile time. Better avoid that at all c=
ost.


>
> Also, before that, I had crash problem when using the heavily
> overloaded PopupMenu::Insertitem() with id and index.  I will send you
> a precise report if this is indeed a problem and not a

Yes, please, that would be nice. You may want to "use Qt::debug;" also to=
 spot=20
the ambiguity (or even better Qt::debug qw| calls ambiguous verbose |)

> misunderstanding from me. I will test where lie the problem by writing
> the same thing in C++.

OK
Cheers,
Germain

>
>
> Thx for your support