[Development] ICU and Windows

[Koehne Kai]

Hi,

Importing ICU into qtbase is fine with me. Anyhow, I don't particular like hard dependency to ICU from Qt5Core, since it forces everyone deploying a Windows helloworld to also ship icudt49.dll with 17,5 MB (!).

I know that you can still configure with -no-icu. So how about making the usage of ICU a runtime decision? Has anybody looked into this already? And are we up to maintain the old, internal backend also in future versions?

Regards

Kai

> -----Original Message-----
> From: development-bounces+kai.koehne=digia.com at qt-project.org
> [mailto:development-bounces+kai.koehne=digia.com at qt-project.org] On
> Behalf Of Thiago Macieira
> Sent: Monday, January 14, 2013 4:52 PM
> To: development at qt-project.org
> Subject: Re: [Development] ICU and Windows
> 
> On segunda-feira, 14 de janeiro de 2013 13.02.46, Shaw Andy wrote:
> > Therefore I would like to propose that for 5.0.1 we simply modify the
> > pro file so that it expects a d after the library name for the debug
> > version and the release one stays as it is.  What we could do to make
> > it more robust is connect it into configure so it checks if it exists
> > and if it does not fall back onto the release version (and give a
> > warning) so it will continue to build as before.
> >
> > Then in 5.1.0 we put ICU into the 3rdparty directory and then we have
> > more control over it and build it ourselves as it seems that this
> > would give us more benefits long term from what John Layt said in a
> previous mail.
> >
> > How does this sound, is there anything that would mean that this is
> > not a good thing to do?
> 
> I think it's too late for 5.0.1. We could do it for 5.0.2, but I'll insist that we
> don't change anything for 5.0.x, unless it is proven that we are doing things
> wrong.
> 
> Let's do the import into 3rdparty for 5.1.0 then, if that's the solution we agree
> upon. And Pau is right: if we need to access the C++ API to get enough
> information for some of our APIs, we'll need to build ourselves for MinGW
> anyway.