Re: Octave 2.9.0 available for ftp

[David Bateman]

John W. Eaton wrote:

> On 17-Mar-2005, David Bateman <address@hidden> wrote:
>
> | Ok. However I just tried to build 2.9.0 if the UMFPACK stuff was built 
> | with the atlas cblas binding. In that case config.log contained messages 
> | like
> | 
> | configure:7268: gcc -o conftest -g -O2   conftest.c -lumfpack  -lamd 
> | -lhdf5 -lz -lm  >&5
> | /usr/lib/gcc/i586-mandrake-linux-gnu/3.4.1/../../../libumfpack.so: 
> | undefined reference to `cblas_dgemv'
> | /usr/lib/gcc/i586-mandrake-linux-gnu/3.4.1/../../../libumfpack.so: 
> | undefined reference to `cblas_dtrsm'
> | /usr/lib/gcc/i586-mandrake-linux-gnu/3.4.1/../../../libumfpack.so: 
> | undefined reference to `cblas_zgemv'
> | 
> | etc. So it seems the test as defined in configure.in is not generic 
> | enough. If there an autoconf macro that just tests for the presence of a 
> | function (nm like functionality) without actually building anything? Is 
> | would avoid this issue...
>
> How about also testing for cblas before testing fro umfpack?
>
> jwe
>
>
>  

What I would suggest that the initial test is left as is. However if 
umfpack is not found then it would be tested with cblas as well, and if 
fond cblas would be added to the umfpack libraries... The build of 
umfpack seems to be really a bit of a mess...

D.


--
David Bateman                                address@hidden
Motorola Labs - Paris                        +33 1 69 35 48 04 (Ph) 
Parc Les Algorithmes, Commune de St Aubin    +33 1 69 35 77 01 (Fax) 
91193 Gif-Sur-Yvette FRANCE

The information contained in this communication has been classified as: 

[x] General Business Information 
[ ] Motorola Internal Use Only 
[ ] Motorola Confidential Proprietary