Re: Building Octave Statically (and without graphics)

[Jason Riedy]

And John W. Eaton writes:
> Building static binaries isn't normal usage these days, so it is not
> surprising to me that some things don't work correctly.

Alas, it is expected usage for certain clusters and parallel
machines.  There is a page detailing how people built an older
Octave for BG/L that *may* be of use to the original poster:
  
http://www.pdc.kth.se/resources/computers/bluegene/bgloctave/mpitb-octave-on-bgl

There would be much (some?) rejoicing if someone would tackle a
--build-static-octave configure option...

> [...] but I don't see a good general way of avoiding the libblas
> xerbla problem.  If I understand correctly, the intent is that users
> should be able to replace that function with their own.

You mentioned the intended method:
>   ar -d libblas.a xerbla.o

We (mostly me) haven't had the mental cycles to figure out the
"right" way to handle XERBLA, and it's made much, much worse by
requirements in the BLAS Tech Forum's BLAS interface (that no one
implements yet).

In my opinion, there likely will be a libxerbla at some future
point so we can avoid cross-library dependencies.  But that's
just my view.

Jason