Re: gdk-pixbuf to GTK
- From: "Shawn T . Amundson" <amundson eventloop com>
- To: gtk-devel-list redhat com
- Subject: Re: gdk-pixbuf to GTK
- Date: Thu, 23 Mar 2000 18:56:23 -0600
On Thu, Mar 23, 2000 at 07:26:24PM -0500, email@example.com wrote:
> > We shouldn't have to maintain tiff/jpeg/png code - someone else already
> > does this.
> *shouldn't* have to, no argument there. But you cannot depend on having
> 1) The appropriate libraries/headers, (especially on embedded systems) or
> 2) The appropriate _versions_ of the libraries/headers
> available on a given machine.
That is why we use autoconf.
> Not to mention that I'd bet dollars to donuts that the interfaces to what
> amounts to the same kind of data (pixmaps) are completely different in all
> three libraries. You're going to have to encapsulate them anyway.
Abstraction code already exists in gdk_pixbuf.
> > We also shouldn't muck around installing things from the GTK+
> > tarball that don't belong there.
> If by this you mean including the current lib[foo] tarballs in with the GTK
> tarballs, you're absolutely correct. Trying to interface the GTK
> configure/make process with these alien libraries would not only be
> painful, but would be difficult to make reliable across multiple platforms
> and multiple compilers/cross compilers.
> That's not what I meant. Instead, I'm talking about extracting the relevent
> code from the image format libraries, and writing native GTK functions that
> implement them.
> Yes, that means every time one of these libs changes that someone has to
> roll the changes into GTK code - but these libraries don't change all that
> often. And the convienince of grabbing GTK and knowing all the appropriate
> bits are included is a big win.
That defeats the entire purpose of packaging up code in library format.
It would also be forking the code for no reason whatsoever.
> > Most people who compile from source probably already have these things
> > because other things do use them.
> You cannot assume that - I've just been through that experience on Solaris.
> And what about Win32 or embedded systems where the libraries don't exist or
> won't compile cleanly as provided by the library maintainers?
All these libraries we are talking about are very well tested on Solaris;
in fact, that's probably where a couple of them were written. And we are
talking about libraries which are optional.
Shawn T. Amundson firstname.lastname@example.org
Research and Development http://www.eventloop.com/
EventLoop, Inc. http://www.snorfle.net/
"The assumption that the universe looks the same in every
direction is clearly not true in reality." - Stephen Hawking
] [Thread Prev