Re: Release Plans

[Daniel J Sebald]

On 07/05/2013 02:42 PM, John W. Eaton wrote:
> On 04/22/2013 03:31 PM, John W. Eaton wrote:
>
> > I would like to aim for making a new stable release by the end of
> > June (this year).
[snip]
> The GUI is better, but we don't have QtHandles integrated and the
> terminal widget does not work properly on either Windows or Unixy
> systems. The biggest problem is that using Ctrl-C for interrupting
> the Octave interpreter does not work and we seem to be unable to make
> keyboard shortcuts work properly, or to allow users to customize them.
> I intend to start a separate thread about these issues. As I see it,
> the GUI is the single largest issue preventing the release from
> happening.

I believe the Ctrl-C issue might tie in with the signals conundrum, 
which has shown up in at least a couple other places.  Right now, Octave 
core will take control of a number of signals, but to do so means 
overriding the GUI's use of signals which happens on at least one 
platform (KDE) when opening files.  The GUI hangs, then eventually 
reports some bad message.  Either signals are going to have to all go to 
the GUI and be routed to the core, or vice versa.

There was a short list of criticals a few weeks back (to which I will 
add one, breakpoints):

1) MinGW/MXE build
2) Bug #38785
3) SIGCHILD (and possibly other) signal handlers (Patch #7949 and Bug 
#38305)
4) Java support
5) breakpoint retention for poorly parsed file

My feeling is that users of the debugger won't find the breakpoint 
behavior acceptable.  The problem is that running a script function for 
which there is a badly constructed line (very easy thing to do) loses 
all breakpoints internally.  Yet, the GUI doesn't update the breakpoint 
markers so it appears that they are still set.  Fix the syntax on the 
offending line and the function will execute all the way through, not 
breaking where previously there were breakpoints.  So one has to go 
through the script file, reapplying breakpoints.  However, the presence 
of the breakpoint markers makes doing so somewhat dodgy.  I think I've 
even managed a scenario where I've saved a file with a bogus line and 
have the function work fine once, then fail on the second attempt.

Dan