Re: Is this a memory leak?

[John W. Eaton]

On 07/19/2016 12:46 PM, Susi Lehtola wrote:
> On 07/19/2016 05:27 AM, John W. Eaton wrote:
> > On 07/19/2016 02:42 AM, Susi Lehtola wrote:
> > > Is this a memory leak, since y is dynamically allocated but it's given
> > > as a dereferenced pointer to octave_value?
> >
> > Yes, it is a leak.  You don't need to create the RowVector object with
> > new.  The memory for RowVector, octave_value, and nearly all other
> > objects in Octave is managed by reference counting.
>
> Right, just as I thought.
>
> > > I tried running octave through valgrind, but I don't get any output..
> >
> > What options did you use for valgrind?
>
> $ valgrind --leak-check=full octave --eval "pkg load
> gsl;A=legendre_sphPlm_array(10,10,0.0);"
> ==6971== Memcheck, a memory error detector
> ==6971== Copyright (C) 2002-2015, and GNU GPL'd, by Julian Seward et al.
> ==6971== Using Valgrind-3.11.0 and LibVEX; rerun with -h for copyright info
> ==6971== Command: octave --eval pkg\ load\
> gsl;A=legendre_sphPlm_array(10,10,0.0);
> ==6971==
> octave: X11 DISPLAY environment variable not set
> octave: disabling GUI features
>
> Without the semicolon you see the array, but again the usual valgrind
> output is missing..?

The "octave" program is now just a small wrapper program that forks and 
execs octave-gui or octave-cli.  So valgrind is just reporting about 
leaks in that program.  Try your test again using octave-cli instead of 
octave.

jwe