Re: [Texmacs-dev] alpha-transparency

[Gubinelli Massimiliano]

As far as understand the main goal is a smooth transition from X11 to  
a modern GUI library (which for the moment is Qt). We proceed by  
perfecting the qt plugins with minimal changes to the core texmacs in  
such a way to guarantee the coexistence of the two interfaces. This  
will allow a smooth transition from the current codebase which is  
quite stable and well tuned (wrt interactive usage) to a more modern  
one based on current GUI toolkits ( Qt, Cocoa, wx, wathever). Indeed  
the main difficulty is the fact that the current codebase is not  
really designed to interact well with modern frameworks (event loop, I/ 
O interaction, standard dialogs, etc...). there are many assumptions  
which are easy to guarantee with X11 (or maybe glut or similar low  
level libraries) but almost impossible to achieve in any reasonable   
modern or cross-platform envirnoment (one for all is that texmacs has  
a very precise idea of the sequencing of operations: input, update,  
redrawing) which is quite difficult to emulate with natural Qt code.  
many of the problems we encountered in developing the qt interface are  
linked to the more or less implicit  assumptions we broke in devising  
the qt version of some X11 constructs. this is still a not very  
pleasant activity, for example we still have some problems at the  
level of socket interaction with plugins. the adantage of this  
conservative approach is that we have always a reliable base of  
comparison. I think that soon (  winter?) we will start to modify  
texmacs internals to remove some of these implicit assumptions which  
hinders developement of more idiomatic qt code (the same assumptions  
however block development of a nice cocoa interface) maybe by  
branching the original codebase. at this moment we will start to  
accept not to be backward compatible with x11 and more interesting  
features will become amenables to addition.

as far as the choice of toolkit : i'm quite satisfied with qt even if  
its behaviour is not quite uniform on different platform when used in  
the quite unorthodox way texmacs forces us to do. maybe it would be  
interesting to experiment with wx, surely i'm personally interested in  
a Cocoa/gnustep port. but for the moment and given the scarcity of  
programming resources we dispose of, consolidation of the qt port  
remains the only reasonable objective to pursue on the short term.

programming texmacs can be frustrating sometimes but the final product  
is good and useful enough that one has seldom the impression of  
loosing its own time.


best
max

On 23 juil. 2010, at 15:59, Norbert Nemec <address@hidden>  
wrote:

> I'm not sure whether Qt is considered "stable" enough. But then -  
> the number of nuissances in the X11 GUI is certainly larger than the  
> number of bugs in the Qt GUI.
>
> I don't see any reason to use Qt exclusively. There are several GUIs  
> in different states of maturity that are worth keeping. It is only  
> X11 that is so limited that I would suggest fading it out.
>
>
> On 23/07/10 13:44, Alvaro Tejero Cantero wrote:
> > Is there any reason not to use the QT GUI everywhere? Any  
> > architecture
> > that supports X11 but not QT?
> >
> > -á.
> >
> >
> >
> > On Fri, Jul 23, 2010 at 11:33, Norbert Nemec<address@hidden 
> > >  wrote:
> >
> > > Hi there,
> > >
> > > it was mentioned before that alpha-transparency should be  
> > > addressed in
> > > TeXmacs. Apart from X11, all other GUIs should support this in a
> > > straightforward way. Is there any idea, how it could be added to  
> > > the GUI
> > > abstraction in the cleanest way so that TeXmacs graphics could  
> > > begin making
> > > use of it?
> > >
> > > Specifically, it would be essential for the table border hints  
> > > that I
> > > attempted to implement a while ago. Currently, the hints are opaque
> > > light-gray and get in the way for equations where the content  
> > > often extends
> > > over the cell boundaries. Putting the hints in the background  
> > > behind the
> > > content was not a solution either, because nested tables would  
> > > then cover
> > > the hints of inner tables by the background color of outer tables.
> > >
> > > The most difficult issue for extending the GUI abstraction seems  
> > > to be a
> > > fallback solution for the X11 GUI and possibly other that do not  
> > > offer alpha
> > > transparency. Once alpha transparency is used for anything  
> > > essential in
> > > TeXmacs, these GUIs would become unusable.
> > >
> > > Is it time to drop the X11 GUI? Or at least accept that future  
> > > versions will
> > > make it less and less usable as alpha transparency is being used  
> > > for more
> > > and more features in TeXmacs? It could still be included as legacy  
> > > code, but
> > > I believe it will slow down the development of TeXmacs in general  
> > > if every
> > > piece of code is restricted to the limitations of the X11 GUI.
> > >
> > > Greetings,
> > > Norbert
> > >
> > > _______________________________________________
> > > Texmacs-dev mailing list
> > > address@hidden
> > > http://lists.gnu.org/mailman/listinfo/texmacs-dev
> > _______________________________________________
> > Texmacs-dev mailing list
> > address@hidden
> > http://lists.gnu.org/mailman/listinfo/texmacs-dev
>
>
> _______________________________________________
> Texmacs-dev mailing list
> address@hidden
> http://lists.gnu.org/mailman/listinfo/texmacs-dev