Re: Please confirm bug #46466

[Julien Bect]

Le 19/05/2016 20:13, Ben Abbott a écrit :

On May 19, 2016, at 13:45, Ben Abbott <address@hidden> wrote:

On May 19, 2016, at 07:01, Ben Abbott <address@hidden> wrote:

On May 19, 2016, at 6:10 AM, Julien Bect <address@hidden> wrote:

Hello all,

Could someone please try to run the example that I have uploaded in comment #4 of bug #46466 ?

@++
Julien

It runs for me on Mac OS 10.10.5 with Octave-4.0.1

Ben

Oops. I didn't realize that the graphic toolkit is related. I used GNUPLOT.

With qt, and the gui, octave hung at iteration 24 on my 1st attempt. On my 2nd attempt it errored out and then hung.

I don't have email via my computer at the moment. If details are needed, I can add that later.

Ben

I typed the error text below

Iter #12/100... error: octave_base_value::convert_to_str_internal (): wrong type argument '<unknown type>'
error: called from
   stk_predict at line 170 column 5
   feval at line 35 column 24
   example at line 17 column 12
error: octave_base_value::convert_to_str_internal (): wrong type argument '<unknown type>'
error: @stk_factorialdesign/size: invalid return value
error: @stk_factorialdesign/size: invalid return value
error: @stk_dataframe/size: invalid return value

Ben

I have found (by bisection) the point on default where the random failures stop to appear (changeset b70cc4bd8109).

This is when JWE started to remove the global_error variable.

I don't understand how exactly this is related to the random failures problem and to Qt, and how to backport a fix to stable.

If you have any idea, please continue this discussion on bug report #46466.

@++
Julien