test failures for chol.cc

classic Classic list List threaded Threaded
5 messages Options
Reply | Threaded
Open this post in threaded view
|

test failures for chol.cc

bpabbott
Administrator
I pulled sources yesterday and built today. I encountered many new  
failures. Most result (151) because I'm missing ARPACK.

I also encountered a chol.cc test failure (lines from fntests.log  
below).

    15 >>>>> processing /Users/bpabbott/Development/mercurial/
octave-3.1.51/src/DLD-FUNCTIONS/chol.cc
    16   ***** test
    17  A = [2,0.2;0.2,1];
    18  issymmetric(A)
    19  min(eig(A))
    20  Ainv = inv(A);
    21  Ainv1 = cholinv(A);
    22  Ainv2 = inv(sparse(A));
    23  Ainv3 = cholinv(sparse(A));
    24  Ainv4 = spcholinv(sparse(A));
    25  assert (norm(Ainv-Ainv1),1e-10)
    26  assert (norm(Ainv-Ainv2),1e-10)
    27  assert (norm(Ainv-Ainv3),1e-10)
    28  assert (norm(Ainv-Ainv4),1e-10)
    29 !!!!! test failed
    30 assert (norm (Ainv - Ainv1),1e-10) expected
    31  1.0000e-10
    32 but got
    33 0

Perhaps, I'm missing the intent, but shouldn't the proper value be  
zero (David)?

    25  assert (norm(Ainv-Ainv1),0,1e-10)
    26  assert (norm(Ainv-Ainv2),0,1e-10)
    27  assert (norm(Ainv-Ainv3),0,1e-10)
    28  assert (norm(Ainv-Ainv4),0,1e-10)

Assuming my understanding is correct, I've attached the trivial  
changeset.

Ben





changeset-chol-tests.patch (1K) Download Attachment
Reply | Threaded
Open this post in threaded view
|

test failures for chol.cc

John W. Eaton
Administrator
On 24-Dec-2008, Ben Abbott wrote:

| I pulled sources yesterday and built today. I encountered many new  
| failures. Most result (151) because I'm missing ARPACK.
|
| I also encountered a chol.cc test failure (lines from fntests.log  
| below).
|
|     15 >>>>> processing /Users/bpabbott/Development/mercurial/
| octave-3.1.51/src/DLD-FUNCTIONS/chol.cc
|     16   ***** test
|     17  A = [2,0.2;0.2,1];
|     18  issymmetric(A)
|     19  min(eig(A))
|     20  Ainv = inv(A);
|     21  Ainv1 = cholinv(A);
|     22  Ainv2 = inv(sparse(A));
|     23  Ainv3 = cholinv(sparse(A));
|     24  Ainv4 = spcholinv(sparse(A));
|     25  assert (norm(Ainv-Ainv1),1e-10)
|     26  assert (norm(Ainv-Ainv2),1e-10)
|     27  assert (norm(Ainv-Ainv3),1e-10)
|     28  assert (norm(Ainv-Ainv4),1e-10)
|     29 !!!!! test failed
|     30 assert (norm (Ainv - Ainv1),1e-10) expected
|     31  1.0000e-10
|     32 but got
|     33 0
|
| Perhaps, I'm missing the intent, but shouldn't the proper value be  
| zero (David)?
|
|     25  assert (norm(Ainv-Ainv1),0,1e-10)
|     26  assert (norm(Ainv-Ainv2),0,1e-10)
|     27  assert (norm(Ainv-Ainv3),0,1e-10)
|     28  assert (norm(Ainv-Ainv4),0,1e-10)
|
| Assuming my understanding is correct, I've attached the trivial  
| changeset.

I applied this changeset.

Thanks,

jwe
Reply | Threaded
Open this post in threaded view
|

test failures for chol.cc

John W. Eaton
Administrator
In reply to this post by bpabbott
On 24-Dec-2008, Ben Abbott wrote:

| I pulled sources yesterday and built today. I encountered many new  
| failures. Most result (151) because I'm missing ARPACK.

I checked in a change that should skip the tests if ARPACK is not
available.

Thanks,

jwe
Reply | Threaded
Open this post in threaded view
|

Re: test failures for chol.cc [changesets for eigs.cc and svds.m]

bpabbott
Administrator

On Dec 24, 2008, at 2:51 PM, John W. Eaton wrote:

> On 24-Dec-2008, Ben Abbott wrote:
>
> | I pulled sources yesterday and built today. I encountered many new
> | failures. Most result (151) because I'm missing ARPACK.
>
> I checked in a change that should skip the tests if ARPACK is not
> available.
>
> Thanks,
>
> jwe
Four "asserts" slipped past.

I assume the same can/should be done for svds.m?

I've attached two changesets.

Ben











changeset-eigs.patch (4K) Download Attachment
changeset-svds.patch (1K) Download Attachment
Reply | Threaded
Open this post in threaded view
|

Re: test failures for chol.cc [changesets for eigs.cc and svds.m]

bpabbott
Administrator

On Dec 25, 2008, at 8:22 PM, Ben Abbott wrote:

> On Dec 24, 2008, at 2:51 PM, John W. Eaton wrote:
>
>> On 24-Dec-2008, Ben Abbott wrote:
>>
>> | I pulled sources yesterday and built today. I encountered many new
>> | failures. Most result (151) because I'm missing ARPACK.
>>
>> I checked in a change that should skip the tests if ARPACK is not
>> available.
>>
>> Thanks,
>>
>> jwe
>
> Four "asserts" slipped past.
>
> I assume the same can/should be done for svds.m?
>
> I've attached two changesets.
>
> Ben

eigs.cc was recently modified so I added "testif HAVE_ARPACK" to 12  
asserts in the current eigs.cc as well as to the three tests in svds,  
and pushed each to Savannah.

Ben