For your information, I ran the tests suite to see how Octave/MSVC behaves.
I got 46 failures out of 3160 tests, and all of these failures were somehow
expected.
1) 1 failure in cellfun.cc
This is due to using fileparts on paths using forward slashes, while fileparts
uses filesep, which is the backslash under MSVC
2) 2 failures in dispatch.cc
These are due to the "echo" command with single quotes. Under Win32
command prompt, the single quotes are echoed as well; hence the generated
dispatch_x.m file is not a valid octave m-file.
3) 4 failures in rand.cc
These are due to the assert calls with absolute tolerance, while a relative
tolerance would be more appropriate when dealing with largs values
(in this case, ~1e9)
4) 26 failures in test_sparse.m
Here I suspect my (highly) hacked version of CXSparse. As MSVC does
not support the C99 _Complex attribute, I converted the whole CXSparse
to C++ using std::complex instead. This is quite a big change and I'm not even
sure that the resulting library is OK.
5) 13 failures in test_system.m
These are due to things not existing under MSVC, like getgrent, getpwnam...
Michael.
[Prev in Thread]
Current Thread
[Next in Thread]
Octave/MSVC regression tests (+ bug in rand.cc),
michael . goffioul<=