[Kde-bindings] Building the QtRuby extension with -export-dynamic

[Richard Dale]

On Wednesday 29 September 2004 19:57, Han Holl wrote:
> On Wednesday 29 September 2004 18:54, Richard Dale wrote:
> > Great! I didn't realise that. So who actually needs to use qtrubyinit and
> > krubyinit executables I wonder. Alex used to have a problem, and other
> > people had crashes in pthreads just before loading the QtRuby extension.
> > Maybe it's been fixed in the latest Qt libs.
> > _______________________________________________
>
> Hi,
>
> I'm running a RH9 system, with Qt-3.1.1, and I can run just fine with
> ruby scriptname, but qtrubyinit works as well.
> This certainly isn't the latest Qt libs.
> (Build Date: Wed 29 Jan 2003 05:04:09 PM CET)
I'm encouraged that non-PowerPC platforms seem to work fine with 
QtRuby/Korundum. I suspect the problem is to do with 'cutting edge' type 
people upgrading their gcc compiler without upgrading the accompanying C++ 
libraries. The trouble is people like that who want the latest gcc, are 
exactly the same people who want to try out language bindings. I had a very 
similar problem with the java bindings where dynamic casts would fail 
randomly. 

About 50% of build problem reports with QtRuby are to do with the qtrubyinit 
executable. The configure check doesn't look for libruby.so correctly, and so 
they get an error, I have to tell people to build ruby with the 
'-enable-shared' option. Or Navis's 'undefined symbol: rb_eRuntimeError' 
problem - I don't know what that is, but he wouldn't have it, if he wasn't 
trying to build qtrubyinit, which he doesn't need.

Hmm.. I'm reluctant to pull qtrubyinit and krubyinit from the release until we 
can nail down what is going on. It would be nice to put rpm/debian/ruby 
gem/Mac OS X .dmg etc binaries on the Rubyforge site, so trying out 
QtRuby/Korundum programming would be more foolproof.

-- Richard