[opensource-dev] LLMatrix3::orthogonalize test (was: LL_TESTS for standalone out-of-source builds)

[Boroondas Gupte]

 On 10/13/2010 03:42 AM, WolfPup Lowenhar wrote:
>
> On the subject of the reason for killing test here is another reason
> that I have attached this is from a full build I did today. The error
> listed in the file is actually coded around and in visual studio you
> have to set that test to not treat warnings as errors so that it will
> not fail!
>
I tried removing the skip line to re-enable the test, like so:

    *diff -r 27535365bd4c indra/llmath/tests/m3math_test.cpp*
    --- a/indra/llmath/tests/m3math_test.cpp	Fri Oct 08 01:13:23 2010 +0200
    +++ b/indra/llmath/tests/m3math_test.cpp	Wed Oct 13 13:16:30 2010 +0200
    @@ -280,7 +280,6 @@
     		llmat_obj.setRows(llvec1, llvec2, llvec3);
     		llmat_obj.orthogonalize();
     
    -		skip("Grr, LLMatrix3::orthogonalize test is failing.  Has it ever worked?");
     		ensure("LLMatrix3::orthogonalize failed ",
     		       is_approx_equal(0.19611613f, llmat_obj.mMatrix[0][0]) &&
     		       is_approx_equal(0.78446454f, llmat_obj.mMatrix[0][1]) &&

On my system that test seems to work fine after that:

    LD_LIBRARY_PATH += ['/home/das-g/slsrc/hg/build/build/llcommon', '/usr/lib', '/usr/local/lib']
    LD_LIBRARY_PATH = '/usr/local/lib:/home/das-g/slsrc/hg/build/build/llcommon:/usr/lib'
    Running: /home/das-g/slsrc/hg/build/build/llmath/INTEGRATION_TEST_m3math
    Unit test group_started name=m3math_h
    Unit test group_completed name=m3math_h
    	Total Tests:	13
    	Passed Tests:	13	YAY!! \o/

Maybe someone has fixed LLMatrix3::orthogonalize but forgot to re-enable
the test? Or is LLMatrix3::orthogonalize behaving differently on
different platforms? (That should affect quaternion calculations and
auto-leveling in Flycam mode, so it's unlikely to go unnoticed.)

Anyway, would be good if others could re-enable the test, too, and see
whether it also passes for them.

Cheers,
Boroondas

PS: LLMatrix3::orthogonalize is a bit of a misnomer. Without having
looked at its code (just at input and desired output in the test) it
seems to orthogonalize *and* normalize (or "ortho*normalize*" for
short). Should the method be renamed to reflect this?
-------------- next part --------------
An HTML attachment was scrubbed...
URL: http://lists.secondlife.com/pipermail/opensource-dev/attachments/20101013/dc1f7d20/attachment.htm