help-octave
[Top][All Lists]
Advanced

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: Memory exhausted when using \


From: Mathieu Dubois
Subject: Re: Memory exhausted when using \
Date: Fri, 20 Jul 2012 22:21:13 +0200 (CEST)


----- Mail original -----
> De: "Jordi Gutiérrez Hermoso" <address@hidden>
> À: "Mathieu Dubois" <address@hidden>
> Cc: "Octave-help" <address@hidden>
> Envoyé: Vendredi 20 Juillet 2012 16:28:37
> Objet: Re: Memory exhausted when using \
>
> On 20 July 2012 08:15, Mathieu Dubois
> <address@hidden> wrote:
>
> >> De: "Jordi Gutiérrez Hermoso" <address@hidden>
>
> >> On 19 July 2012 16:17, Jordi Gutiérrez Hermoso
> >> <address@hidden>
> >> wrote:
> >> > On 19 July 2012 16:06, Mathieu Dubois
> >> > <address@hidden>
> >> > wrote:
> >> >> You can download the dump at:
> >> >> http://www.ensta-paristech.fr/~mdubois/memory-exhausted
> >> >>
> >> >>
> >> >> The command I used was:
> >> >> save memory-exhausted A b
> >> >> right before the line:
> >> >>
> >> >> tic; X = A \ b; toc;
> >> >> (executing this line raises the error)
> >> >
> >> > Ok, great, thanks. I can now reproduce this problem, and I think
> >> > it
> >> > shouldn't happen. This may be a bug in Octave. I'm investigating
> >> > now.
> >>
> >> I can see the problem now. I just patched it:
> >>
> >>     http://hg.savannah.gnu.org/hgweb/octave/rev/4315a39da4c9
> >>
> >> You'll have to wait for our 3.6.3 release or apply this patch
> >> yourself
> >> to the current 3.6.2 release.
> >>
> > I have tried to patch octave 3.6.2 but it fails altough
> > differently.
> > I'm not completely sure if this is correct.
>
> Yes, from your description, you patched Octave correctly.
>
> > When I run test_inpaintZ it stops on the first test (the one with
> > the small matrices). The output message and the core dump are
> > attached.
>
> I'm unable to reproduce this problem with the code you've sent me,
> but
> I was also unable to reproduce the "memory exhausted" error message
> to
> begin with. Either you're sending me code that is different from the
> code that you're running yourself, or... well, I don't know what else
> could be happening. I am suspicious of the nvidia blob that shows up
> in your memory map, but that could be unrelated.
>
> If you want to keep debugging this, do the following:
>
>     1) Compile Octave again, this time as a debug build:
>
>        ./configure CFLAGS=-g CXXFLAGS=-g FFLAGS=-g
>        make
>        ## no make install
>
>    2) Run Octave with gdb, from the build directory:
>
>      ./run-octave -g /path/to/test_inpaintZ.m
>      (gdb) run
>
>    3) When Octave crashes, get a stack trace:
>
>      (gdb) bt
>
>    4) Send us that stack trace.
>
> You should also do this again but with valgrind instead of gdb
> (-valgrind option to ./run-octave instead of -g).

I have followed your instructions (except that I copied the necessary files in 
build directory so that Octave finds them). I ran "make distclean" before 
calling "./configure CFLAGS=-g CXXFLAGS=-g FFLAGS=-g".

Octave seems to crash at the first test with the memory exhausted error (the 
command was executed before A\b; the file is attached).
When I run bt, gdb answers "No stack." (it says "[Inferior 1 (process 12079) 
exited normally]" after Octave message).
This is really puzzling. As you mentioned you can't reproduce the bug...

For the code except if I made a mistake you have the same code. Anyway, I have 
attached a new copy.

If that helps, I'm under Ubuntu 12.04 on a 64 bits platform (gcc 4.6.3, kernel 
3.2.0-26-generic).

>
> > Is that a problem?
>
> Well, of course, Octave shouldn't crash.
>
> - Jordi G. H.
>

Attachment: inpaintZ_octave.tar.gz
Description: application/compressed-tar

Attachment: memory-exhausted
Description: Binary data


reply via email to

[Prev in Thread] Current Thread [Next in Thread]