bugGNU Octave - Bugs: bug #63384, Error for "inv(NaN)"...

 
 

bug #63384: Error for "inv(NaN)" with LAPACK 3.11.0

Submitter:  Markus Mützel <mmuetzel>
Submitted:  Fri 18 Nov 2022 06:14:04 PM UTC
   
 
Category:  Test Suite Severity:  3 - Normal
Priority:  5 - Normal Item Group:  Incorrect Result
Status:  Fixed Assigned to:  None
Originator Name:  Open/Closed:  * Closed
Release:  * dev Operating System:  * Any
Fixed Release:  8.1.0 Planned Release:  8.1.0
* Mandatory Fields

Add a New Comment Rich Markup
   

Jump to the original submission

Mon 05 Jun 2023 05:34:14 PM UTC, comment #41: 
Dmitri A. Sergatskov <dasergatskov>
Mon 05 Jun 2023 05:17:06 PM UTC, comment #40: 

Fedora's lapack-3.11-04 includes the latest lapack's "fix" for this issue and with it I see:


octave:1> test libinterp/corefcn/inv.cc-tst
***** warning <rcond = > assert (inv (Inf (2,2)), NaN (2,2))
!!!!! warning failed.
Expected warning <rcond = >, but got error <ASSERT errors for:  assert (inv (Inf (2, 2)),NaN (2, 2))

  Location  |  Observed  |  Expected  |  Reason
   (1,1)         Inf          NaN        'NaN' mismatch
   (2,1)         Inf          NaN        'NaN' mismatch
   (1,2)         Inf          NaN        'NaN' mismatch
   (2,2)         Inf          NaN        'NaN' mismatch>


and with the test program:


$ gfortran bug-63384.f -llapack -lblas
$ ./a.out
 x =
                       NaN                       NaN
                       NaN                       NaN
 anorm =                        NaN
 dgetrf info:            0
 ipiv =
           1           2
 ** On entry to DGECON parameter number  5 had an illegal value
Note: The following floating-point exceptions are signalling: IEEE_INVALID_FLAG


Dmitri.
--

Dmitri A. Sergatskov <dasergatskov>
Tue 29 Nov 2022 08:04:57 PM UTC, comment #39: 

Marking as Fixed and closing report.

Rik <rik5>
Group administrator
Tue 29 Nov 2022 08:03:49 PM UTC, comment #38: 

The error part of this bug report has been fixed.  I filed another report (bug #63445) about the difference in the calculated inverse.

Rik <rik5>
Group administrator
Thu 24 Nov 2022 06:01:59 PM UTC, comment #37: 

The additional output in Octave (rcond) is an extension and doesn't affect Matlab compatibility so we are fine there.  The question for me was whether the calculated inverse should match Matlab.  The results seem more reasonable (NaN for Matlab instead of Inf for Octave).

Rik <rik5>
Group administrator
Thu 24 Nov 2022 03:17:49 PM UTC, comment #36: 

Fwiw, in Matlab R2022a:

>> x = nan (2,2)

x =

   NaN   NaN
   NaN   NaN

>> [y,rc] = inv (x)
Error using inv
Too many output arguments.

>> y = inv (x)
Warning: Matrix is singular, close to singular or badly scaled. Results may be inaccurate. RCOND = NaN.


y =

   NaN   NaN
   NaN   NaN

>> inv (single (x))
Warning: Matrix is singular, close to singular or badly scaled. Results may be inaccurate. RCOND = NaN.


ans =

  2×2 single matrix

   NaN   NaN
   NaN   NaN

>> inv (complex (single (x), ones (2,2)))
Warning: Matrix is singular, close to singular or badly scaled. Results may be inaccurate. RCOND = NaN.


ans =

  2×2 single matrix

      NaN +    NaNi      NaN +    NaNi
      NaN +    NaNi      NaN +    NaNi

>> inv (complex (x, ones (2,2)))
Warning: Matrix is singular, close to singular or badly scaled. Results may be inaccurate. RCOND = NaN.


ans =

      NaN +    NaNi      NaN +    NaNi
      NaN +    NaNi      NaN +    NaNi

>> version -lapack

ans =

    'Intel(R) oneAPI Math Kernel Library Version 2021.3-Product Build 20210611 for Intel(R) 64 architecture applications (CNR branch auto) supporting Linear Algebra PACKage (LAPACK 3.9.0)'

>> version -blas

ans =

    'Intel(R) oneAPI Math Kernel Library Version 2021.3-Product Build 20210611 for Intel(R) 64 architecture applications (CNR branch auto)'


It looks like they don't return the inverse condition number. But since they do not return anything else either, this is probably an extension we can keep without worrying too much about compatibility.
Other than that, their results make more sense to me...

Markus Mützel <mmuetzel>
Group administrator
Thu 24 Nov 2022 02:41:31 PM UTC, comment #35: 

Should we bother with Matlab compatibility here?  Or maybe this is a different issue.  See test code below:


octave:1> x = nan (2,2)
x =

   NaN   NaN
   NaN   NaN

octave:2> [y,rc] = inv (x)
y =

   NaN   NaN
   NaN   NaN

rc = NaN
octave:3> [y,rc] = inv (single (x))
y =

   NaN     0
   NaN   NaN

rc = NaN
octave:4> [y,rc] = inv (complex (single (x), ones (2,2)))
y =

   Inf   Inf
   Inf   Inf

rc = 0
octave:5> [y,rc] = inv (complex (x, ones (2,2)))
y =

   Inf   Inf
   Inf   Inf

rc = 0


In Matlab, the return value is always all NaN for every entry.  Complex matrices are returned with "NaN + NaNi" for each element.

Rik <rik5>
Group administrator
Wed 23 Nov 2022 03:05:57 PM UTC, comment #34: 

This is a good question. We have not yet found why the problem shows up in 3.11 and not in 3.10. There were some changes in LATRS ( https://github.com/Reference-LAPACK/lapack/pull/712 ) and that might be a side effect. As of now, we do not know why the problem shows up in 3.11 and not in 3.10.

Julien Langou <langou>
Wed 23 Nov 2022 02:56:06 PM UTC, comment #33: 

Thanks for the pointer.
Just out of interest and to better estimate if other parts of our code could be affected: Did you find out which change to which function caused the regression between LAPACK 3.10 and 3.11?

Markus Mützel <mmuetzel>
Group administrator
Wed 23 Nov 2022 02:53:56 PM UTC, comment #32: 
Julien Langou <langou>
Tue 22 Nov 2022 07:40:10 PM UTC, comment #31: 

Oops. Pushed the changes to the files for complex valued matrices here:
https://hg.savannah.gnu.org/hgweb/octave/rev/ec973756cb3a

Markus Mützel <mmuetzel>
Group administrator
Mon 21 Nov 2022 05:37:03 PM UTC, comment #30: 

comment #29:

> Is there a similar problem for complex-valued arrays?


Yes, both CMatrix.cc and fCMatrix.cc have the function finverse which uses the same problematic coding construct.

Rik <rik5>
Group administrator
Mon 21 Nov 2022 05:26:55 PM UTC, comment #29: 

Is there a similar problem for complex-valued arrays?

John W. Eaton <jwe>
Group administrator
Mon 21 Nov 2022 05:23:33 PM UTC, comment #28: 

I pushed a slightly different version of the patch to the stable branch here:
https://hg.savannah.gnu.org/hgweb/octave/rev/55f0173e4f08

For symmetry reasons, I made the same change for the FloatMatrix version of the function.

Marking as ready for test.

Markus Mützel <mmuetzel>
Group administrator
Mon 21 Nov 2022 03:33:18 PM UTC, comment #27: 

The patch helps:



octave:1> version -lapack
ans = Linear Algebra PACKage Version 3.11.0
octave:2> test libinterp/corefcn/inv.cc-tst
PASSES 62 out of 62 tests



Dmitri.
--

Dmitri A. Sergatskov <dasergatskov>
Mon 21 Nov 2022 03:25:30 PM UTC, comment #26: 

I submitted an issue:

 https://github.com/Reference-LAPACK/lapack/issues/763

Just to see what they say.

Dmitri.
--

Dmitri A. Sergatskov <dasergatskov>
Mon 21 Nov 2022 03:16:43 PM UTC, comment #25: 

Does the following change help?

diff -r 288a917d728e liboctave/array/dMatrix.cc
--- a/liboctave/array/dMatrix.cc        Sun Nov 20 15:29:28 2022 +0100
+++ b/liboctave/array/dMatrix.cc        Mon Nov 21 15:48:48 2022 +0100
@@ -28,6 +28,7 @@
 #endif

 #include <algorithm>
+#include <cmath>
 #include <istream>
 #include <limits>
 #include <ostream>
@@ -596,19 +597,24 @@
     info = -1;
   else if (calc_cond)
     {
-      F77_INT dgecon_info = 0;
-
-      // Now calculate the condition number for non-singular matrix.
-      char job = '1';
-      Array<F77_INT> iz (dim_vector (nc, 1));
-      F77_INT *piz = iz.fortran_vec ();
-      F77_XFCN (dgecon, DGECON, (F77_CONST_CHAR_ARG2 (&job, 1),
-                                 nc, tmp_data, nr, anorm,
-                                 rcon, pz, piz, dgecon_info
-                                 F77_CHAR_ARG_LEN (1)));
-
-      if (dgecon_info != 0)
-        info = -1;
+      if (std::isnan (anorm))
+        rcon = std::nan ("0");
+      else
+        {
+          F77_INT dgecon_info = 0;
+
+          // Now calculate the condition number for non-singular matrix.
+          char job = '1';
+          Array<F77_INT> iz (dim_vector (nc, 1));
+          F77_INT *piz = iz.fortran_vec ();
+          F77_XFCN (dgecon, DGECON, (F77_CONST_CHAR_ARG2 (&job, 1),
+                                     nc, tmp_data, nr, anorm,
+                                     rcon, pz, piz, dgecon_info
+                                     F77_CHAR_ARG_LEN (1)));
+
+          if (dgecon_info != 0)
+            info = -1;
+        }
     }

   if (info == -1 && ! force)


Markus Mützel <mmuetzel>
Group administrator
Mon 21 Nov 2022 03:06:53 PM UTC, comment #24: 

"Anyway, we could probably add a short-hand in our code if `anorm` is NaN here (and not proceed to use LAPACK at all)"

That seems like a good solution to me.

John W. Eaton <jwe>
Group administrator
Mon 21 Nov 2022 02:38:30 PM UTC, comment #23: 

Is it possible to stop in Fortran code with a debugger?
If it is, what is the value of AINVNM in both versions of LAPACK in dgecon.f:250?
https://github.com/Reference-LAPACK/lapack/blob/master/SRC/dgecon.f#L250

Maybe, it was finite before and is NaN in the newer version?

Anyway, we could probably add a short-hand in our code if `anorm` is NaN here (and not proceed to use LAPACK at all):
https://hg.savannah.gnu.org/hgweb/octave/file/288a917d728e/liboctave/array/dMatrix.cc#l587

Markus Mützel <mmuetzel>
Group administrator
Mon 21 Nov 2022 01:57:24 PM UTC, comment #22: 

Maybe, this is caused by "rcond" now being returned as 0 (instead of NaN) together with this in our code:

      if ((calc_cond || mattype.ishermitian ()) && rcon == 0.0)
        ret = Matrix (rows (), columns (),
                      octave::numeric_limits<double>::Inf ());


The documentation for dgecon says about rcond:
https://netlib.org/lapack/explore-html/dd/d9a/group__double_g_ecomputational_ga188b8d30443d14b1a3f7f8331d87ae60.html

>           RCOND is DOUBLE PRECISION
>           The reciprocal of the condition number of the matrix A,
>           computed as RCOND = 1/(norm(A) * norm(inv(A))).


So, rcond should be NaN in this case if I understand correctly...

Markus Mützel <mmuetzel>
Group administrator
Mon 21 Nov 2022 01:44:50 PM UTC, comment #21: 

Also:


$ gfortran -O2 -march=native -mtune=native bug-63384.f  -lflexiblas
$ ./a.out
 x =
                       NaN                       NaN
                       NaN                       NaN
 anorm =                        NaN
 dgetrf info:            0
 ipiv =
           1           2
 dgecon info:            0
 rcond =                        NaN
 dgetri info:            0
 xinv =
                       NaN                       NaN
                       NaN                       NaN
$ LD_PRELOAD=/usr/lib64/liblapack.so ./a.out
 x =
                       NaN                       NaN
                       NaN                       NaN
 anorm =                        NaN
 dgetrf info:            0
 ipiv =
           1           2
 dgecon info:            0
 rcond =    0.0000000000000000
 dgetri info:            0
 xinv =
                       NaN                       NaN
                       NaN                       NaN
$ ldd a.out
        linux-vdso.so.1 (0x00007fff94a65000)
        libflexiblas.so.3 => /lib64/libflexiblas.so.3 (0x00007f8ebe400000)
        libgfortran.so.5 => /lib64/libgfortran.so.5 (0x00007f8ebe000000)
        libm.so.6 => /lib64/libm.so.6 (0x00007f8ebe324000)
        libgcc_s.so.1 => /lib64/libgcc_s.so.1 (0x00007f8ebe864000)
        libquadmath.so.0 => /lib64/libquadmath.so.0 (0x00007f8ebe81a000)
        libc.so.6 => /lib64/libc.so.6 (0x00007f8ebdc00000)
        /lib64/ld-linux-x86-64.so.2 (0x00007f8ebe8a3000)


Dmitri.
--



Dmitri A. Sergatskov <dasergatskov>
Mon 21 Nov 2022 01:40:55 PM UTC, comment #20: 


$ ./a.out
 x =
                       NaN                       NaN
                       NaN                       NaN
 anorm =                        NaN
 dgetrf info:            0
 ipiv =
           1           2
 dgecon info:            0
 rcond =    0.0000000000000000
 dgetri info:            0
 xinv =
                       NaN                       NaN
                       NaN                       NaN

$ ldd a.out
        linux-vdso.so.1 (0x00007fff37bdd000)
        liblapack.so.3 => /lib64/liblapack.so.3 (0x00007fee75200000)
        libblas.so.3 => /lib64/libblas.so.3 (0x00007fee759b2000)
        libgfortran.so.5 => /lib64/libgfortran.so.5 (0x00007fee74e00000)
        libm.so.6 => /lib64/libm.so.6 (0x00007fee75124000)
        libgcc_s.so.1 => /lib64/libgcc_s.so.1 (0x00007fee75997000)
        libquadmath.so.0 => /lib64/libquadmath.so.0 (0x00007fee750da000)
        libc.so.6 => /lib64/libc.so.6 (0x00007fee74a00000)
        /lib64/ld-linux-x86-64.so.2 (0x00007fee75a5f000)


$ rpm -qf /usr/lib64/liblapack.so.3
lapack-3.11.0-1.fc35.x86_64


Dmitri.
--

Dmitri A. Sergatskov <dasergatskov>
Mon 21 Nov 2022 01:25:52 PM UTC, comment #19: 

Try the attached file.  Can you reproduce the problem with it?  Compile with


gfortran bug-63384.f  -llapack -lblas


(file #53980)

John W. Eaton <jwe>
Group administrator
Mon 21 Nov 2022 11:57:16 AM UTC, comment #18: 

Thanks. IIUC, that conclusively shows that this is a regression with LAPACK 3.11.0 (on any platform). Re-titling accordingly.

Unfortunately, I can't write Fortran. But I agree that it might be good to ask upstream if they also think this is a regression (or falls into the "undefined behavior" category).

Markus Mützel <mmuetzel>
Group administrator
Mon 21 Nov 2022 11:51:31 AM UTC, comment #17: 


$ ./run-octave -qf
octave:1> version -lapack
ans = Linear Algebra PACKage Version 3.10.1
octave:2> version -blas
ans = FlexiBLAS Version 3.2.1
OpenBLAS (config: unknown)
octave:3> test libinterp/corefcn/inv.cc-tst
PASSES 62 out of 62 tests
octave:4>

$ LD_PRELOAD=/usr/lib64/liblapack.so ./run-octave -qf
octave:1> version -lapack
ans = Linear Algebra PACKage Version 3.11.0
octave:2> version -blas
ans = FlexiBLAS Version 3.2.1
OpenBLAS (config: unknown)
octave:3> test libinterp/corefcn/inv.cc-tst
***** warning <rcond = > assert (inv (NaN (2,2)), NaN (2,2))
!!!!! warning failed.
Expected warning <rcond = >, but got error <ASSERT errors for:  assert (inv (NaN (2, 2)),NaN (2, 2))

  Location  |  Observed  |  Expected  |  Reason
   (1,1)         Inf          NaN        'NaN' mismatch
   (2,1)         Inf          NaN        'NaN' mismatch
   (1,2)         Inf          NaN        'NaN' mismatch
   (2,2)         Inf          NaN        'NaN' mismatch>

octave:4>


Openblas is 0.3.21

Dmitri.
--

Dmitri A. Sergatskov <dasergatskov>
Mon 21 Nov 2022 07:20:04 AM UTC, comment #16: 

@Dmitri: It's still not clear to me which combination of BLAS and LAPACK "flavors" and versions was used in the two tests in comment #11.
Could you please also execute `version -blas` and `version -lapack` for the two different LD_PRELOAD?
Does it work correctly if you don't LD_PRELOAD anything?

Markus Mützel <mmuetzel>
Group administrator
Sun 20 Nov 2022 04:22:11 PM UTC, comment #15: 

FYI --
There are two tests failing with lapack-3.11 (inv and xinv):


processing /home/dima/src/octave/clang_opt/libinterp/corefcn/inv.cc-tst
***** warning <rcond = > assert (inv (NaN (2,2)), NaN (2,2))
!!!!! warning failed.
Expected warning <rcond = >, but got error <ASSERT errors for:  assert (inv (NaN (2, 2)),NaN (2, 2))

  Location  |  Observed  |  Expected  |  Reason
   (1,1)         Inf          NaN        'NaN' mismatch
   (2,1)         Inf          NaN        'NaN' mismatch
   (1,2)         Inf          NaN        'NaN' mismatch
   (2,2)         Inf          NaN        'NaN' mismatch>

***** test
 [xinv, rcond] = inv (NaN (2,2));
 assert (xinv, NaN (2,2));
 assert (rcond, NaN);
!!!!! test failed
ASSERT errors for:  assert (xinv,NaN (2, 2))

  Location  |  Observed  |  Expected  |  Reason
   (1,1)         Inf          NaN        'NaN' mismatch
   (2,1)         Inf          NaN        'NaN' mismatch
   (1,2)         Inf          NaN        'NaN' mismatch
   (2,2)         Inf          NaN        'NaN' mismatch


If somebody can write a fortran code that reproduces this test, it would help to push it upstream.

Dmitri.
--

Dmitri A. Sergatskov <dasergatskov>
Sun 20 Nov 2022 04:05:14 PM UTC, comment #14: 

(And by "compile" I meant "linked statically with")

Dmitri.
--

Dmitri A. Sergatskov <dasergatskov>
Sun 20 Nov 2022 04:01:15 PM UTC, comment #13: 

By default it linked to openblas-0.3.21 (which I assume was compiled against lapack 3.10 ).

Dmitri.
--

Dmitri A. Sergatskov <dasergatskov>
Sun 20 Nov 2022 03:53:49 PM UTC, comment #12: 

@Dmitri: Which BLAS libraries were used in your first test? And which LAPACK libraries were used in your second test?

Markus Mützel <mmuetzel>
Group administrator
Sun 20 Nov 2022 03:50:01 PM UTC, comment #11: 

On Fedora with lapack 3.11:


$ LD_PRELOAD=/usr/lib64/liblapack.so ./run-octave -qf
octave:1> test libinterp/corefcn/inv.cc-tst
***** warning <rcond = > assert (inv (NaN (2,2)), NaN (2,2))
!!!!! warning failed.
Expected warning <rcond = >, but got error <ASSERT errors for:  assert (inv (NaN (2, 2)),NaN (2, 2))

  Location  |  Observed  |  Expected  |  Reason
   (1,1)         Inf          NaN        'NaN' mismatch
   (2,1)         Inf          NaN        'NaN' mismatch
   (1,2)         Inf          NaN        'NaN' mismatch
   (2,2)         Inf          NaN        'NaN' mismatch>


but:


$ LD_PRELOAD=/usr/lib64/libblas.so ./run-octave -qf
octave:1> test libinterp/corefcn/inv.cc-tst
PASSES 62 out of 62 tests
octave:2>



Dmitri.
--

Dmitri A. Sergatskov <dasergatskov>
Sun 20 Nov 2022 03:34:12 PM UTC, comment #10: 

We 'shouldnt' be

John Donoghue <lostbard>
Group Member
Sun 20 Nov 2022 03:27:59 PM UTC, comment #9: 

Are we using the LAPACK functions from OpenBLAS when we switch BLAS libraries?

Markus Mützel <mmuetzel>
Group administrator
Sun 20 Nov 2022 03:26:44 PM UTC, comment #8: 

I downloaded the .7z archive from https://buildbot.octave.space/#/download and extracted it locally.

The test is also failing with reference LAPACK:

>> version -blas
ans = unknown or reference BLAS
>> version -lapack
ans = Linear Algebra PACKage Version 3.11.0
>> inv (NaN (2,2))
warning: matrix singular to machine precision
ans =

   Inf   Inf
   Inf   Inf


If I'm following the code correctly, the inversion of a double precision, full, non-complex matrix should be calculated by `Matrix::finverse`:
https://hg.savannah.gnu.org/hgweb/octave/file/288a917d728e/liboctave/array/dMatrix.cc#l550

Maybe, a regression in LAPACK 3.11?

Markus Mützel <mmuetzel>
Group administrator
Sun 20 Nov 2022 03:07:42 PM UTC, comment #7: 

I meant that if we can run it without the failure using ref blas, then it would indicate that the issue may indeed be within openblas

John Donoghue <lostbard>
Group Member
Sun 20 Nov 2022 01:45:44 PM UTC, comment #6: 

We are using OpenBLAS for those tests (because that is the default when using the installer):
https://github.com/gnu-octave/octave-buildbot/blob/main/.github/workflows/check.yaml#L111

Markus Mützel <mmuetzel>
Group administrator
Sun 20 Nov 2022 01:42:43 PM UTC, comment #5: 

If openblas, we should be able to set the default blas to the reference one, which I dont think has changed


John Donoghue <lostbard>
Group Member
Sun 20 Nov 2022 01:05:59 PM UTC, comment #4: 

The patches for bug #49091 should only be affecting the result of inverse trigonometric or hyperbolic functions (e.g., acosh and alike). I'm not sure if those are used for matrix inversion.

Maybe a linear algebra library? Like OpenBLAS or alike?

Markus Mützel <mmuetzel>
Group administrator
Sun 20 Nov 2022 12:54:12 PM UTC, comment #3: 

Default has the Bug #49091 patches etc that I dont think the release version has

John Donoghue <lostbard>
Group Member
Sun 20 Nov 2022 12:45:41 PM UTC, comment #2: 

I have a octave 7.3.1 build using the default branch using commit in mxe of

b06843747eca add opkg-cfitsio


That works - so it either started happening after that? That was long other gcc and mingw-w64 updates that I thought were possible culprits


John Donoghue <lostbard>
Group Member
Fri 18 Nov 2022 06:43:25 PM UTC, comment #1: 

Possible cultits I guess are:

 gcc- v11 - v12

 mingw-w64 v9 - v10

 
And some differences in patches for the above packages


John Donoghue <lostbard>
Group Member
Fri 18 Nov 2022 06:14:04 PM UTC, original submission:  

Some tests in "inv.cc-tst" are failing for the stable branch of Octave with the default branch of MXE Octave:
https://github.com/gnu-octave/octave-buildbot/actions/runs/3498715372/jobs/5859432776#step:11:86480

>>>>> processing D:\a\OCTAVE~1\OCTAVE~1\OCTAVE~1\mingw64\share\octave\8.0.1\etc\tests\libinterp\corefcn\inv.cc-tst
***** warning <rcond = > assert (inv (NaN (2,2)), NaN (2,2))
!!!!! warning failed.
Expected warning <rcond = >, but got error <ASSERT errors for:  assert (inv (NaN (2, 2)),NaN (2, 2))

  Location  |  Observed  |  Expected  |  Reason
   (1,1)         Inf          NaN        'NaN' mismatch
   (2,1)         Inf          NaN        'NaN' mismatch
   (1,2)         Inf          NaN        'NaN' mismatch
   (2,2)         Inf          NaN        'NaN' mismatch>

***** test
 [xinv, rcond] = inv (NaN (2,2));
 assert (xinv, NaN (2,2));
 assert (rcond, NaN);
!!!!! test failed
ASSERT errors for:  assert (xinv,NaN (2, 2))

  Location  |  Observed  |  Expected  |  Reason
   (1,1)         Inf          NaN        'NaN' mismatch
   (2,1)         Inf          NaN        'NaN' mismatch
   (1,2)         Inf          NaN        'NaN' mismatch
   (2,2)         Inf          NaN        'NaN' mismatch


Those tests are succeeding for the stable branch of Octave with the release branch of MXE Octave. So, this is likely a regression with one of the dependencies that have been updated since the release of Octave 7.

CC'ing John Donoghue because I hope he might have a hunch as to who might be the "culprit".

Markus Mützel <mmuetzel>
Group administrator

 

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

Attach Files:
   
   
Comment:
   

Attached Files
file #53980:  bug-63384.f added by jwe (934B - text/x-fortran)

 

Depends on the following items: None found

Items that depend on this one: None found

 

Carbon-Copy List
  • -email is unavailable- added by langou (Posted a comment)
  • -email is unavailable- added by rik5 (Posted a comment)
  • -email is unavailable- added by jwe (Updated the item)
  • -email is unavailable- added by dasergatskov (Posted a comment)
  • -email is unavailable- added by lostbard (Posted a comment)
  • -email is unavailable- added by mmuetzel (Submitted the item)
  • -email is unavailable- added by mmuetzel
  •  

    There are 0 votes so far. Votes easily highlight which items people would like to see resolved in priority, independently of the priority of the item set by tracker managers.

    Only group members can vote.

     

    Follow 10 latest changes.

    Date Changed by Updated Field Previous Value => Replaced by
    2022-11-29 rik5 Open/ClosedOpen Closed
    2022-11-29 rik5 StatusReady For Test Fixed
        Fixed ReleaseNone 8.1.0
        Planned ReleaseNone 8.1.0
    2022-11-21 mmuetzel StatusConfirmed Ready For Test
    2022-11-21 jwe Attached File- Added bug-63384.f, #53980
    2022-11-21 mmuetzel StatusNone Confirmed
        Operating SystemMicrosoft Windows Any
        Summary[MXE Octave] Error for &quot;inv&quot; with default branch Error for "inv(NaN)" with LAPACK 3.11.0
    2022-11-18 mmuetzel Carbon-Copy- Added lostbard

    Back to the top

    Powered by Savane 3.13-02a9.
    Corresponding source code