Re: the latest LAPACK

[John W. Eaton]

On 22-Jun-1999, Ross A. Lippert <address@hidden> wrote:

| It is LAPACK barfing.

Unless Octave is using LAPACK incorrectly, I'd have to say that the
bug is definitely in LAPACK, not Octave.

| However, I'd like you to consider this LAPACK
| bug as one reason why an update to LAPACK 3.0 might benefit octave,

I agree.  Once it is released, I have no problem with upgrading.  I
did that shortly after LAPACK 2.0 was released.  Yours was the first
message that I have seen about any new version of LAPACK.

| bc at the user level it is octave which barfs.

But I don't think it is fair to blame Octave.  First, I don't remember
anyone reporting a problem or suggesting a way that Octave could work
around the bug in LAPACK.  If Octave misbehaved because of a bug in
the C library, would you say that Octave was `barfing'?  Or would you
say, `the bug is in the C library; fix it and the problem will go away'?

| I will try out your svd example once I have succeeded in building
| LAPACK 3.0.  I just got it.

| I pointed out a bug in LAPACK last
| year and got a good response and a patch, but that was after I had
| traced the bug to the few lines of fortran code in LAPACK which were
| causing the problem.

Well, I don't have time for that.  If I did, I could probably just fix
the problem myself.

How often do I see much more than "it's broke! fix it!"?

jwe