Ticket #14796 (pike): please commit

[Guido Soranzio]

On Apr 1, 2008, at 6:46 AM, Jordan K. Hubbard wrote:

> Well, let's also be careful not to conflate multiple, desirable  
> goals here - that only increases the size of the task and the  
> attendant risk of seeing nothing happen.

My initial propose was very simple and pragmatic, indeed.

We don't have a central building machine, we don't have binary
packages, we have difficulties in testing universal packages.
But we have an "archive mode", we can produce packages,
we can temporarily create a symbolic link to switch off
/opt/local/sandbox.

If you are a commiter or even a simple user that has spent
hours while building a MacPorts subtree from scratch,
please share it with us, so we can speed up the commits
by the currently few volunteers. Do you have the latest
Monodevelop compiled for Quartz or Carbon, an advanced
scientific software packaged from SourceForge into /opt/local,
an optimized Blender 3D? Please upload your archives so
we can commit the new patches starting from there.

I had to build for hours the latest LAMP server, GCC,
boost, qt4-mac, GStremer codecs, WebKit, gnome 2.22 from
scratch only in order to test the uncountable untested broken
ports. It took me days to build them after assuring that
I was using the latest available versions: other users/committers
could make use of my GBs of precompiled binaries.

Benjamin Reed from the Fink project is offering his 3 GB
builds of KDE by using BitTorrent: that's a simple and effective
initiative based on DMGs which even developers can benefits
from.

We have already the well-made free Porticus GUI and we have already
the new sqlite receipt database engine in Leopard's Apple
InstallerPackageMaker: we only need pkgutils to fully support
--unlink, reference counting and dependency analysis. But even
if pkgutil was hosted on MacOSForge, probably it would ignore
MacPorts like the Xquartz, ZFS, MacRuby and WebKit projects are
doing now, even if they are technically real "Mac ports"!


- Guido