Performance issues on Windows, suggests a MSVC build

[John W. Eaton]

On 28-Jun-2011, Thomas Weber wrote:

| On 27-Jun-2011, John W. Eaton wrote:
| > On 27-Jun-2011, Thomas Weber wrote:
| > > Eh, no. I don't think that a soname of 0.0.0 is a sign of a properly
| > > handled soname. This is not an issue in itself right now, but soname
| > > handling isn't that easy - I can't tell you much about it due to lack
| > > of knowledge on my part. But as I said, it doesn't matter much - it's
| > > a non-issue right now.
| 
| > What number should we have used as a starting point?
| 
| Right, that shall teach my not to jump to conclusions. I thought the
| 0.0.0 just happens to be added by libtool automatically, not that it was
| the actual start for the soname numbering.
| 
| However, I think that the 3.2 series did the following by default:
|       ln -s liboctave.so liboctave.so.3.2.4
| If that is true (can't check right now), I suggest starting with
| something that is bigger than 3.2.4 as soname.

The libtool manual clearly states that the numbering scheme for
sonames should not correspond in any way to the version of the
package, as that tends to cause confusion.  We could start with 4.0.0,
but I don't see that as being any better than 0.0.0.  And in any case,
we already released a version of Octave with the library versions set
to 0.0.0.

jwe