Re: [AD] Allegro 4.2.0: hsv_to_rgb needless assert

[Peter Wang <tjaden@xxxxxxxxxx>]

On 2005-12-21, Elias Pschernig <elias@xxxxxxxxxx> wrote:
> On Wed, 2005-12-21 at 09:50 +0000, Vincent Penecherc'h wrote:
> > > Ah, yes, only backwards compatibility. Very good.
> > 
> > With the same soname ?
> >
> > Not that I particularly care, just asking mostly to have
> > the point under consideration.

It seems to be the lesser of two evils.  If you bump the soname due to
forwards incompatibility you force some users to upgrade needlessly, but
if you don't bump the soname then binaries linked against newer versions
may load against older versions of the library and only break
(mysteriously) at run time.  Hopefully we won't have many
forwards-incompatible changes.

Allegro also confounds the library version number with the soname, which
is a no-no.  Luckily we define the 4.2 branch as "those versions of the
library which are [backwards] compatible with previous 4.2.x releases"
so I think that choice is, at least, justified :-)

> They are named liballeg-4.2.0.so and liballeg-4.2.1.so. And a symlink
> liballeg.so.4.2 is created, which points to liballeg-4.2.1.so - and
> executables apparently use that symlink to find the actual lib.. not
> sure how that works.

The liballeg.so.4.2 comes from the soname.

Peter