[Changeset]: Re: More "float" problems in default branch for Mac

[John W. Eaton]

On 23-May-2008, David Bateman wrote:

| David Bateman wrote:
| > Thomas Treichl wrote:
| >   
| >> David Bateman schrieb:
| >>     
| >>> Thomas Treichl wrote:
| >>> Ok, I now have a changeset to fix this issue, but there was a secondary
| >>> issue that prevented compilation.. You'll need the two changesets
| >>> attached..
| >>>
| >>> Sorry
| >>> David
| >>>       
| >> No problem, now it's me who must say that I'm sorry for being
| >> annoying: Your patches brought me through balance.cc but now I'm
| >> failing at lu.cc, ideas?
| >>
| >>   /usr/bin/ld: Undefined symbols:
| >>   base_lu<FloatComplexMatrix, std::complex<float>, Matrix,
| >> double>::L() const
| >>   base_lu<FloatComplexMatrix, std::complex<float>, Matrix,
| >> double>::P() const
| >>   base_lu<FloatComplexMatrix, std::complex<float>, Matrix,
| >> double>::U() const
| >>   base_lu<FloatComplexMatrix, std::complex<float>, Matrix,
| >> double>::P_vec()
| >>     const
| >>   collect2: ld returned 1 exit status
| >>   make[2]: *** [lu.oct] Error 1
| >>   make[1]: *** [src] Error 2
| >>   make: *** [all] Error 2
| >>
| >> I've commented out compilation of lu.cc, then I succeeded, so what I
| >> mean is that this seems to be the last problem that still is there. I
| >> can now work on the test fails on Mac for the default branch,
| >> hopefully solving some on my own ;)
| >>
| >>   Thomas
| >>
| >>     
| > Opps, again my fault.. It shows that I haven't gotten the test code for
| > this written yet, and you are turning up a number of bugs.. Try the
| > attached changeset
| >
| > D.
| >
| >   
| >
| 
| Humm, too fast.. The permutation vector should always be of double type
| even for the single precision LU, as its effectively storing the
| octave_idx_type values. So the attached is more correct.

I applied this changeset.

Thanks,

jwe