Re: Successfully merged projects

[John Swensen]

On Apr 11, 2011, at 4:52 AM, Michael Goffioul wrote:

> On Mon, Apr 11, 2011 at 5:19 AM, John Swensen <address@hidden> wrote:
>> I see that you cut out the MDI interface.  In the meantime, I was working on 
>> a simple file browser dockable panel.  If you want to pull the updated 
>> FileChooserWidget.h and FileChooserWidget.cpp into your source tree, then I 
>> can start working off you GitHub project exclusively (now I have to learn 
>> git, despite my best efforts to avoid it ;)).  Here are a few other things 
>> that I think we will want to do:
>> 1) In the MainWindow::establishOctaveLink (by the way, I think your change 
>> to the OctaveLink class name is much more descriptive and apropos than 
>> octave_server), you want to put a small sleep between the calls to 
>> pthread_create and the command_editor::add_event_hook.  On OSX it was 
>> crashing because Octave hadn't been initialized by the time add_event_hook 
>> was called.  Maybe there is a variable in Octave that gets set when 
>> initialization is complete and we can wait for that to be set, rather than 
>> an arbitrary sleep time that could fail under heavy system load conditions.
>> 2) I think you want to add the definition of an extern for the quit_allowed 
>> boolean and set it to true somewhere after octave_main has been called.  
>> There is some weirdness about how we need to start octave_main in embedded 
>> mode, switch to buffer, then run the octave main_loop manually.  As a result 
>> of things that happen inside octave_main when the embedded flag is set to 
>> one, we have to trick Octave into allowing us to call exit from the command 
>> line in the GUI.  A better solution would be to propose a patch to Octave 
>> that allows the third parameter of octave_main to specify NORMAL, EMBEDDED, 
>> and IDE modes and set things up appropriately.
>> extern bool quit_allowed;
>> quit_allowed = true; // Called after octave_main
> 
> This is not portable to Windows. That variable is defined in toplev.h,
> why can't you just include that header?
> Also, why can't you just call octave_main with embedded=false?
> 
> Michael.

I think having quit_allowed exposed in toplev.h as an extern is a new thing.  
I'm pretty sure it used to be a variable only available in toplev.cc.  The 
reason we need to have it run in embedded mode is to prevent 
clean_up_and_exit() from being called when the main_loop finishes.  For various 
reasons, mostly associated with interacting with Octave through the (now 
called) OctaveLink class, the UI should control when Octave is cleaned up.  
That is why I said that it would be preferable to have the third parameter to 
octave_main be an enum indicating the mode of operation and act accordingly, 
rather than just embedded or not embedded.

John Swensen