Re: testing install of suitesparse

[John W. Eaton]

On 24-Oct-2007, Ben Abbott wrote:

| 
| On Oct 24, 2007, at 7:54 PM, John W. Eaton wrote:
| 
| > On 24-Oct-2007, Ben Abbott wrote:
| >
| > | You're correct. I checked the fntests.log and found the ismember.m
| > | did report an error. However, since the proper response was  
| > given, it
| > | appears to be a false failure.
| > |
| > |  >>>>> processing /Users/bpabbott/src/octave/07.10.12/scripts/set/
| > | ismember.m
| > |    ***** xtest assert (ismember ('', {'abc', 'def'}), false);
| > | !!!!! known failure
| > | error: assert (ismember (, {"abc", "def"}),false) expected
| > | 0
| > | but got
| > | [](0x0)
| > | Dimensions don't match
| > |    ***** xtest fail ('ismember ([], {1, 2})', 'error:.*');
| > | !!!!! known failure
| > | error: expected error <error:.*> but got none
| >
| > I don't understand what you mean by "false failure".  In both of the
| > cases above, what ismember actually does is different from what we
| > would like it to do.
| >
| > jwe
| 
| Perhaps my inferred understanding of the test is lacking, but it  
| appears to me that ismember('',{'abc','def'}) produced [](0x0) ... as  
| it should. However, the message indicates an "expected" result of 0.
| 
| This is further complicated by the phrase: "error: expected error  
| <error:.*> but got none" ... reminds me of a professor I had who  
| flunked me because he was certain I wasn't capable of the A's I'd  
| scored on his tests ;-)
| 
| In any event, perhaps, my qualification, "false failure", is  
| improper, but I don't understand why a correct answer to a test  
| produces an error.

The first test is

  xtest assert (ismember ('', {'abc', 'def'}), false);

The "xtest" prefix means "If this test fails, count it as an expected
failure.  If it succeeds, count it as a PASS."

The test is the assert call.  The expected value is the last
parameter, false.  But in this case, there is a bug, and ismember
returns [].  So normally, assert would report a failed test.  But
since this is an xtest, it is counted as an expected failure.

Likewise for the second one above.  We are expecting the function call

  ismember ([], {1, 2})

to throw an error, but because there is a known bug, it does not throw
an error.  Normally this would be reported as a failed test, but since
this is an xtest, it is counted as an expected failure.

Does this make sense now?

Oh, and would someone like to look again at actually fixing these bugs
in ismember?  I tried, but found it to be a bit of a mess of special
cases and I could never get it right.  Maybe someone else would have
better luck.

jwe