Re: fntests errors on mingw-Octave 1:dispatch.cc

[Tatsuro MATSUOKA]

Hello John

--- "John W. Eaton" <address@hidden> wrote:
> I don't think it would be a waste of time to fix these tests so that
> they work properly on Windows and Unixy systems.

OK. I see!!  We have to solve the problems.  
As I have reported, the test under the cygwin was almost complete except for 
rand.cc.
The rand.cc problem has been aalready solved by David. 
The cygwin environments give posix emulation so that it is rather unixy but not 
windows like from the
point of view of the fntests.    

We have to come back to Michael's first mail which is cited at the end of the 
main for convinience. 
This is because the failures on mingw are similar to those by Michael. 

The differences are described the below:

****************
Mingw intrinsic;
>  octave-3.0.0\src/ov-fcn-handle.cc  PASS    2/3    FAIL 1
  ***** test
 a = 2;
 f = @(x) a + x;
 g = @(x) 2 * x;
 h = @log2;
 f2 = f;
 g2 = g;
 h2 = h;
 nm = tmpnam();
 unwind_protect
   save ("-text", nm, "f2", "g2", "h2");
   clear f2 g2 h2
   load (nm);
   assert (f(2),f2(2));
   assert (g(2),g2(2));
   assert (g(3),g2(3));
   unlink (nm);
   save ("-text", nm, "f2", "g2", "h2");
 unwind_protect_cleanup
   unlink (nm);
 end_unwind_protect
!!!!! test failed
error: function handle points to non-existent function

>  + test_io.m freezes during the test.
***************
MSVC intrinsic
***************
While
 test_string - 1 failure


********************
We have to discuss the common problems and the intrinsic problems
separately.

We begin with the commom problem first.
 1.dispatch.cc - 2 failures
 2.general/bitcmp.m - 1 failure
 3.test_system - 14 failures

Oh! I forget to discuss that it is allowable the 'fntests' have platform 
dependent parts.
Otherwise nothing goes forward.

I'll wait the replies.

Regards

Tatsuro



Michel's mail
===============================================

The package compiles fine under MSVC. I ran the test suite and
here are the results:

dispatch.cc - 2 failures
These are due to the fact the argument to
"system" is not run into a shell, so something like
system("echo 'something'>file.out") does not generate file.out; hence
the dispatch tests that use this trick to produce temporary m-file fail.

general/bitcmp.m - 1 failure
 ***** assert(bitcmp(A,Amax),bitor(bitshift(1,Amax-1),bitshift(1,Amax-2)));
!!!!! test failed
error: assert (bitcmp (A, Amax),bitor (bitshift (1, Amax - 1),
bitshift (1, Amax - 2))) expected
6.7554e+015
but got
1.1259e+016
values do not match
shared variables {
 Amax =  53
 Bmax = 9.0072e+015
 A = 2.2518e+015
}

test_string - 1 failure
This is due to the fact that under MSVC: isprint(setstr(9)) == 1,
while the test expects 0.

test_system - 14 failures
The reasons are various:
- stat(filesep) fails; what works is stat('\\') or stat('C:\')
- getpwuid, getprgp, getppid, getpwent, getgrent are not implemented
- doing "cd /" and "pwd ()" does not give you "/", but something
including the current drive letter (like "d:\")

In summary, almost all failures are "expected" under MSVC. The only
one that could be problematic is the one related to bitcmp. However,
as I'm no expert, I don't really understand what's going on.




--------------------------------------
Easy + Joy + Powerful = Yahoo! Bookmarks x Toolbar
http://pr.mail.yahoo.co.jp/toolbar/