Re: [Texmacs-dev] Cocoa backend

[Massimiliano Gubinelli]

On 16 Dec 2006, at 15:34, Felix Breuer wrote:

> On Sat, 2006-12-16 at 22:13 +0100, Massimiliano Gubinelli wrote:
> > Hi,
> >   thank you all for the comments.
[snip]
> You mention replacing the UI widgets. How do you handle the drawing of
> the actual document? AFAIK TeXmacs uses an abstract "PS-device" which
> maps the drawing commands to X11 calls. Is it possible to reuse  
> that one
> on Cocoa? Or did you start implementing a Cocoa backend for the
> PS-device? Would a OpenGL/Cairo backend serve you as well? I think the
> implementation of such an OpenGL or Cairo backend would be a great
> benefit for TeXmacs on all platforms.
>
> Also I think it would be great if TeXmacs were decomposed into a  
> set of
> reusable libraries. GUIs could then be implemented more easily on  
> top of
> these. This would not only increase the maintainability of the GUI but
> also the maintainability of TeXmacs itself. (I have not tried to
> undertake such a decomposition, because I do not have the necessary
> skills.)
>
>
> Good luck with your port,
> Felix
>
>
> PS: There are numerous bugs in TeXmacs with regard to memory  
> management
> as exemplified by the reports on the bug tracker about crashes. Any  
> help
> in fixing those would be greatly appreciated. I can't do much except
> provide you with a number of test-cases.

Writing the MacOSX "PS-device" it has not been too difficult. It has  
been enough to map X11 calls to the analogous calls in Cocoa. To  
reimplement the device in Cairo or directly in OpenGL should be a  
matter of a week or less. Actually what is requiring a lot of work is  
to properly handle events since the Cocoa's event model is slightly  
different from X11's. (you can do a straight translation but the  
result is not good Cocoa coding).

I think that one of my remarks has not been properly appreciated:  
this "port" run ALSO on Linux if one is willing to use GNUstep  
libraries (and on Windows). If the community is pondering a common  
platform-independent API then GNUstep should be considered as a  
valuable candidate which would have the benefit of running "native"  
on MacOSX (via Cocoa which is Apple implementation of the NextStep/ 
OpenStep API on which also GNUstep is based). Moreover on Unix  
machines using Qt, GTK, GNUstep or whatever should be considered an  
equivalent choice (for what I understand).



max

PS: For the moment I'm not interested in memory management but on  
MacOSX there are also tools to detect memory leaks (MallocDebug,  
ObjAlloc...) so if somebody with a Mac would give them a try...