[Biopython-dev] Re: [BioPython] tests failing

[Brad Chapman]

Jeff:
> Thanks for letting us know about these.  I'm moving this thread onto 
> the "biopython-dev" list, as it's probably more appropriate there.

I'd like to second the thanks -- it's all around nice to have people
using biopython regularly on non-UNIX platforms. 

> >Failure: test_SubsMat
> >
> >AssertionError:
> >output: 'M0.00 0.40 0.70 0.80 1.00\n'
> >Expected: 'M -0.00 0.40 0.70 0.80 1.00\n'
> 
> It looks like this is from a difference in how windows and Iddo's OS 
> handles 0's.  It's probably not serious, but should be fixed.  Iddo, 
> can you write some code that will check for this?

I think this actually might be a python version difference and not an
OS difference. I'm also seeing it right now on my machine:

$ uname -a
NetBSD taxus.athen1.ga.home.com 1.5.1 NetBSD 1.5.1 (TAXUS) #1: Tue Jun 12 09:13:48 EDT 2001     chapmanb at taxus:/usr/src/sys/arch/macppc/compile/TAXUS macppc

$ python
Python 2.1 (#6, Jul  8 2001, 17:18:01) 
[GCC egcs-2.91.66 19990314 (egcs-1.1.2 release)] on netbsd1

FAIL: test_SubsMat
----------------------------------------------------------------------
Traceback (most recent call last):
  File "run_tests.py", line 153, in runTest
    expected_handle)
  File "run_tests.py", line 247, in compare_output
    assert expected_line == output_line, \
AssertionError: 
Output  : 'M 0.00 0.40 0.70 0.80 1.00\n'
Expected: 'M -0.00 0.40 0.70 0.80 1.00\n'

This is just one I've thrown my hands up in the air about. It's not
really a bug in SubsMat (hey, 0.00 and -0.00 are still the same, right
:-), but I'm not sure how to make the regression checker recognize this.

> >Failure: test_prodoc
> >
> >AssertionError:
> >Output: 'J. \n'
> >Expected: 'J. \n'
> 
> Brad, this looks pretty odd.  Is it a newline problem?

This is another one I've seen on Windows and also on Yair's Mac stuff,
but have to throw my hands up in the air about. What Mark reported
here is different from what I've seen -- my error looks like:

Output: 'J. \n'
Expected: 'J.\n'

So, there is, for some unknown reason, as extra space generated at the
end of the line, that we don't see on UNIX platforms. I'm not sure
what is going on here, or how we can make the regression tester stop
choking on it (other than reintroducing my "end of the line whitespace
isn't important stuff" :-). 

Any ideas for anyone? I'd definately like to clear up these two
problems if we could.

Brad