bugGNU Octave - Bugs: bug #35596, Some 'make check' tests fail

 
 

bug #35596: Some 'make check' tests fail

Submitted by:  Lluís Batlle <viric>
Submitted on:  Thu 23 Feb 2012 11:33:15 AM UTC  
 
Category: NoneSeverity: 3 - Normal
Priority: 5 - NormalItem Group: Incorrect Result
Status: InvalidAssigned to: None
Originator Name: Open/Closed: Closed
Release: 3.6.0Operating System: GNU/Linux

Add a New Comment(Rich Markup)
   

You are not logged in

Please log in, so followups can be emailed to you.

 

Thu 23 Feb 2012 09:03:37 PM UTC, comment #4:

Oh, this was a lapack 3.4.0 bug. It's in their errata.

http://www.netlib.org/lapack/Errata/index2.html (bug0085)

This can be closed.

Lluís Batlle <viric>
Thu 23 Feb 2012 08:16:34 PM UTC, comment #3:

I don't see the problem with lapack 3.3.1 on my Debian system.

But with lapack 3.4.0, I see the DORGLQ parameter 5 error plus a crash due to an apparent memory bug.

Octave does not call DORGLQ directly. [15:11]

So if you want to help debug this problem, run Octave under a debugger and find the point where DORGLQ is being called incorrectly. Then find out why that is happening.

Is it because Octave is calling some LAPACK function incorrectly, or is it that there is a bug in LAPACK that results in the incorrectly call?

John W. Eaton <jwe>
Project Administrator
Thu 23 Feb 2012 07:42:25 PM UTC, comment #2:

One person using Arch Linux confirms it on his x86_64, too, with this same test. Same outcome.

Lluís Batlle <viric>
Thu 23 Feb 2012 11:42:32 AM UTC, comment #1:

I forgot to mention that the dgesvd parameters m,3,n=5,lda=3,ldvt=1 is causing a call to dorglq with m=5,n=5,lda=lvdt=1.

LDA violates the requirement for dorglq for LDA>=max(1,m): http://www.netlib.no/netlib/lapack/double/dorglq.f

I don't understand the dgesvd requirements enough to know if the call dgesvd is correct, because it's its internal call to dorglq that fails.

http://www.netlib.no/netlib/lapack/double/dgesvd.f

For the gdb backtrace before, I used "break dorglq_" and then run the test in octave.

Lluís Batlle <viric>
Thu 23 Feb 2012 11:33:15 AM UTC, original submission:

I'm trying octave built with liblapack 3.4.0 on GNU/Linux x86_64 and qrupdate 1.1.2. qrupdate, lapack, and atlas tests work all fine. I get the same outcome with octave 3.4.3 and 3.6.0.

Some tests of the octave "make check" fail for me, related to lapack. I don't list them all here.

I pick one of the tests as an example, as at least someone got it working fine in his system, but not me. I run this test:

This fails for me this way:

Building lapack and octave with debug symobls, I could check with gdb where all this happens. Backtrace:

jwe at #octave claimed that the test works for his octave.

Can someone reproduce this?

Lluís Batlle <viric>

 

(Note: upload size limit is set to 16384 kB, after insertion of the required escape characters.)

Attach File(s):
   
   
Comment:
   

No files currently attached

 

Depends on the following items: None found

Digest:
   bug dependencies.

 

Carbon-Copy List
  • -unavailable- added by jordigh (Updated the item)
  • -unavailable- added by jwe (Posted a comment)
  • -unavailable- added by viric (Submitted the item)
  •  

    Do you think this task is very important?
    If so, you can click here to add your encouragement to it.
    This task has 0 encouragements so far.

    Only project members can vote.

     

    Please enter the title of George Orwell's famous dystopian book (it's a date):

     

     

    Follow 3 latest changes.

    Date Changed By Updated Field Previous Value => Replaced By
    Fri 20 Jul 2012 04:26:49 PM UTCjordighDependencies-=>bugs #36904 is dependent
    Thu 23 Feb 2012 09:05:40 PM UTCjordighStatusNone=>Invalid
      Open/ClosedOpen=>Closed

    Back to the top


    Powered by Savane 3.1-cleanup1