[libdvdcss-devel] le grande cleanup

[Diego Biurrun]

On Thu, Oct 30, 2014 at 08:24:18PM +0100, Jean-Baptiste Kempf wrote:
> On 30 Oct, Diego Biurrun wrote :
> > On Thu, Oct 30, 2014 at 02:01:28PM +0100, Jean-Baptiste Kempf wrote:
> > > On 29 Oct, Diego Biurrun wrote :
> > > > I think Windows 9x should go for sure.  The others we can discuss.
> > > 
> > > Be careful: what is called here "Win9x" is in fact Adaptec SCSI
> > > interface and runs also on Windows XP.
> > 
> > I was wondering myself due to the "ASPI" in the names.  That still leaves
> > the question of whether or not this code is worth keeping.
> 
> I'm not sure we should keep it, tbh.
> At worse, it's in the git.

This code cannot have seen much testing in the last 10 years ever since
Windows 9x faded into obsolescence.  Here's a better log message:

  Drop support for ASPI disc access (and thus Windows 9x).

  The ASPI access method is only relevant on Windows 9x versions where no
  alternative exists. However, These Windows versions are obsolete since a
  long time and have no more real world usage. Also, the ASPI code is ugly
  and a maintenance burden.

Good to go then?

> > > > [PATCH 43/46] Drop support for BeOS.
> > > > [PATCH 44/46] Drop support for QNX.
> > > > [PATCH 45/46] Drop support for OS/2.
> > > > [PATCH 46/46] Drop support for Solaris.
> > > 
> > > I think this is not the way to go:
> > > The way to go is to split this code in different files, one per OS, with
> > > a common code calling the OS-specific functions so that supporting the rarest
> > > OSes does not burden the maintenance of the other ones.
> > 
> > I have had this very thought myself.  Maybe I'll get to it.
> 
> It feels cleaner, to me, and as libdvdcss is not that big piece of code,
> it wouldn't hurt too much to keep other OSes in.

I don't mind keeping the other (fringe) OSes, but I'd like to reevaluate
if they work at all, if they are worth keeping, and whether or not all
that support code is truly necessary.

IMO HP-UX is a complete fringe WTF system and Windows 9x just requires too
much clutter, thus both are not worth the trouble of continuing to deal
with them.

I strongly suspect that the BeOS code is no longer necessary for Haiku
and I similarly suspect that half the (Open)BSD hacks are now obsolete.

> But I agree 1000x that the ifdef mess we have must go away.

Nobody can disagree about this I think.

Diego