Re: Fwd: [swig:bugs] #1353 Octave 3.8.0 support

[John W. Eaton]

On 01/07/2014 06:24 AM, Jordi Gutiérrez Hermoso wrote:
> > -------- Original Message --------
> > Subject: [swig:bugs] #1353 Octave 3.8.0 support
> > Date: Sun, 05 Jan 2014 13:45:11 +0000
> > From: William Fulton <address@hidden>
> > Reply-To: [swig:bugs]  <address@hidden>
> > To: [swig:bugs]  <address@hidden>
> >
> > APIs always change even if not intentional, someone will someday
> > mess up and we will need an API version number to fix these kind of
> > errors.
>
> I don't understand. What use is an API version number if two
> unintentionally different APIs have the same version number? Our
> argument is that you can't trust API version numbers at all, which is
> why you test for features, not numbers.

I checked in the following changeset:

http://hg.savannah.gnu.org/hgweb/octave/rev/b6b6e0dc700e

This will be part of the 3.8.1 release.  So if
OCTAVE_API_VERSION_NUMBER is defined, you can do what you did before.
If you have OCTAVE_{MAJOR,MINOR,PATCH}, you can check for features in
3.8.1 and later.  If none of these are defined, then you have 3.8.0.

At least the version number will change with each release.  As Jordi
is trying to tell you, sometimes we have accidentally made changes and
forgotton to update the API version number.  Well, nobody's perfect...

Of course, even the version number won't help with development
versions where pretty much anything goes and we certainly aren't
changing the version number for each hg commit, so I don't know how to
help you there.  But then again, OCTAVE_API_VERSION_NUMBER was not
reliable for development versions either.

jwe