Hi all,
Couple of observations from afar (yes, I am still here watching how my baby is doing :-)
Having dealt with a lot of remote dev, etc before, I would _really_ recommend that you create a stable and experimental CVS branch as Guenter mentions. In past projects I have been involved with, the "active" developers would work on the experminental branch, knowing that at any moment, the entire thing could be broken. "Normal" users or developers would use the stable branch. Every so often, when the experimental branch is good to go, then it replaces the stable branch, and you fork out another experimental one. Also, labeling in CVS is easy and should be used whenever a big checkin happens so people can get back to a known state.
On the issue of the processImage stuff, I totally agree that the original design assumptions I made 7 years ago aren't true anymore :-) Remember that all of that code started on an SGI Indy...machines are a bit faster these days. This is one reason why it is such a pain to do streaming pixel manipulation (like on movies). A complete overhaul of the pix system would be a lot of work, but would probably help a lot.
I have actually been looking into ways to have objects "advertise" their connection abilities and just let them negotiate the best way to run. This allows for optimized processing for known formats, as well as conversion for formats that the original code hasn't even seen before. BTW, I would recommend that YUV stuff is folded into the normal pix flow. While I know that there are significant differences, etc. between the two formats, it is confusing to end users who don't know the differences. If pixel operations are dealt with differently on Mac vs all the other platforms, then it will be harder to transfer patches and general GEM knowledge for most people independent of the platform.
Just my $.02
Later, Mark
============================== = mdanks@stormfront.com = Technical Director PS2/XBox = http://www.danks.org/mark ==============================
-----Original Message----- From: chris clepper [mailto:cclepper@artic.edu] Sent: Monday, December 16, 2002 5:04 PM To: guenter geiger Cc: PD-dev@iem.kug.ac.at Subject: Re: [PD-dev] [GEM] new postrender crashes OS X
A slighlty different proposal. Having taken a closer look at Gem, and specifically pix_ processing, I do completely understand that Johannes wants to change everything in there. There is no way around introducing bugs in this process.
In order to make this transition less painfull, we could
make a stable
branch of Gem, which will only change for bugfixes, build system enhancements etc (no new features !!).
This means of course that there will be no yuv support.
Is there a need for such a branch ?
Interesting proposition Guenter. Maybe a stable and 'experimental' branch could help ease transitions in code additions. Another idea might be to have a 'Core' Gem version that's cross-platform with extra extensions to accommodate differences and unique features for each OS. That might alleviate some of our problems. The counterpoint is that it would mean more work for administrators and potential confusion for users. Some thing to think about...
cgc
Guenter
--
PD-dev mailing list PD-dev@iem.kug.ac.at http://iem.kug.ac.at/cgi-bin/mailman/listinfo/pd-dev
-
Mark Danks