[Octave-bug-tracker] [bug #63410] mean input checking misses dim = Inf, poor error messaging

[Nicholas Jankowski]

URL:
  <https://savannah.gnu.org/bugs/?63410>

                 Summary: mean input checking misses dim = Inf, poor error
messaging
                 Project: GNU Octave
               Submitter: nrjank
               Submitted: Wed 23 Nov 2022 01:42:46 PM EST
                Category: Octave Function
                Severity: 3 - Normal
                Priority: 5 - Normal
              Item Group: Unexpected Error or Warning
                  Status: None
             Assigned to: None
         Originator Name: 
        Originator Email: 
             Open/Closed: Open
                 Release: 8.0.90
         Discussion Lock: Any
        Operating System: Any
           Fixed Release: None
         Planned Release: None


    _______________________________________________________

Follow-up Comments:


-------------------------------------------------------
Date: Wed 23 Nov 2022 01:42:46 PM EST By: Nicholas Jankowski <nrjank>
(not sure which item group this should go in)

as noticed in patch #10288,  mean.m fails to consistently throw an error if
Nan or Inf is passed as a dim input.  

In 7.3.0:

>> mean([1,2,3],NaN)
error: mean: DIM must be an integer and a valid dimension
error: called from
    mean at line 134 column 9

>> mean([1,2,3],Inf)
error: conversion of inf to octave_idx_type value failed
error: called from
    mean at line 150 column 5



testing the mean function from 8.0.90 and 9.1.0 (within 7.3.0, so not certain
the difference isn't due to the mixed environment):

>> mean([1,2,3],NaN)
error: conversion of nan to octave_idx_type value failed
error: called from
    mean at line 166 column 9
>> mean([1,2,3],Inf)
error: conversion of inf to octave_idx_type value failed
error: called from
    mean at line 166 column 9

 
in any case, the type checking for dim should catch both nan and inf.  

Looking at the code, it looks like mean got a lot of rework from 7->8.  In
7.3.0 the following dim check catches Nan but misses Inf:


if (dim != fix (dim) || dim < 1)
        error ("mean: DIM must be an integer and a valid dimension");
      endif


in 8+, the dim check was changed to:
    if (! (isvector (dim) && all (dim)) || any (rem (dim, 1)))
      error ("mean: DIM must be a positive integer scalar or vector");
    endif

which misses both. rem(Inf,1) and rem(NaN,1) both return NaN, and any(NaN) =
0.  

patch #10288 suggested dim!=fix(dim) changed to mod(dim,1)~=0 would catch both
Inf and NaN. I think that would apply here as well.  Will try it and add some
BISTs to see if anything else slips through.







    _______________________________________________________

Reply to this item at:

  <https://savannah.gnu.org/bugs/?63410>

_______________________________________________
Message sent via Savannah
https://savannah.gnu.org/