[fpc-pascal] FPC for FreeBSD AMD64?

[Marco van de Voort]

In our previous episode, Francisco Reyes said:
> >> Also, when one uses  instal.sh, how does one uninstall?
> > 
> > Not in theory. In practice it amounts to 
> > - killing /usr/local/lib/fpc/ and /usr/local/share/fpc.
> >   (or only the relevant version you installed if you have multiples),
> > - weeding out binaries in /usr/local/bin. 
> 
> Any reason not to have an uninstall.sh?

Yes. Nobody made it. Moreover the install.sh doesn't save what files it
stored, and what dirs were (partially?) created. It only decompresses
archives, it has no clue what is in them.

And of course deletion is more dangerous than installation. A mistake in
scripting and you might wipe /.

> In particular install.sh could be changed to save the directories where it 
> was installed and then create the uninstall.sh.
   
That would unlink $PREFIX/bin since it stores binaries there. You really
need to do it on a per file level :-)

> > Florian did a few bits last week, to help me over the point where I got
> > stuck. We are currently investigating what is the next step.
> 
> Would it make sense for a newbie to even try to look at buildfaq or should I 
> wait to see what the "next step" will be?.

Both. Wait as far as x86_64 is concerns, but familiarize yourself with the
buildfaq as much as possible (crosscompile between linux and -64, and
between linux and freebsd). This since even if it gets done, it will be
rough the first months.
 
> Is there a list, other than this, to keep track of the progress of the 
> FreeBSD AMD64 port? 

No. For an update you can also ask on IRC.

> Also, once the port is finished  will it always be up to 
> the same level as the i386 version in FreeBSD?

That depends on your (plural, as in the future FreeBSD/x86_64 users) patches
and bugreports. Too little interest might cause it to preliminary die.

Maintenance costs effort, and needs input. Interest seemed on the rise, so
we gave it a go to do the initial work, but the longterm input and bugfixing
need to come from users.

It is really demotivating if you notice a major bug when you are packaging
release $X, and find out it was in the year old $X-1 release too, and nobody
noticed. That's what happened to the 1.0.x NetBSD and OpenBSD ports.

> > Note that once a port is done, it usually needs a lot of testing and fixing
> > of all kinds of smaller libraries. It might not be directly up to the level
> > of other platforms. 
> 
> Do we have a regression test?

Yes, in directory tests/ in the FPC sources (might not be in release
sources, but a snapshot surely has it). FreeBSD/x86_64 had 26 "run" errors
last weekend. I found a bug in statfs, and will see how it runs tonight.

As soon as it can compile itself, I'll upload a snapshot. Next will be
getting access to C libraries (linking to shared libs)