Re: Problem installing octave-forge packages

[John Swensen]

John Swensen wrote:
> John W. Eaton wrote:
> > On  5-Apr-2007, John Swensen wrote:
> >
> > | I recently installed the most recent CVS and now I can't get 
> > packages to | install.
> >
> > What is recently?
> >
> > | I get a segmentation fault every time.  I dug around for a
> > I can't duplicate this problem.  Do you have multiple copies of Octave
> > installed that might be trying to use the same libraries?
> >
> > jwe
> >
> >   
> I just tried again with the latest CVS as of 4/10/2007 at about 6PM.  
> I am still having the same problem.
> I always install each of my builds to a different directory in 
> /sw/opt, but just to make sure they weren't conflicting, I deleted all 
> of them before starting a new build.  I didn't have a lot of time last 
> night to investigate more, but it is definitely still happening and 
> still occurs in the save_vars() function.  I will try to find time to 
> investigate more later today.
>
> John Swensen
I debugged some more via GDB and narrowed down the problem even further.
All of these line numbers are for code from CVS as of April 13, 2007 at 
~8PM.

In load-save.cc at line 1733, in the DEFCMD(save), there is the 
following function call
   save_vars (argv, i, argc, file, format, save_as_floats, 
write_header_info);
I checked each of these variables in GDB by backing up the call stack.  
Here, i=1.  However, when I go one more step down the call stack, inside 
the save_vars function, the function parameter (argv_idx) that 
corresponds to the input variable 'i' is -1073753728. 

I am not a GDB expert, but it seems strange that all of the variables 
before and after this function call have the same values, except 
i/argv_idx.  Any suggestions on how to proceed?

John Swensen