sev wrote:
Let me explain it as a 0.8.0 release coordinator.
MacOS X is one of the main development platforms in ScummVM. Several key developers and moreover our co-lead Fingolfin run it every day. So it is very tested. Moreover, there is no difference between MacOS, Windows, Linux, *nix, OS/2 ports. All they use exactly same codebase (except some Altivec code in MacOS and some audio drivers). That means that if there is a bug in MacOS, it will appear in Windows too (technically endianness could play a role here, but every engine was tested on both big- and little-endian platforms, and only sword2 had alignment ptoblems which are already fixed in 0.9.0). All of that makes MacOS test builds redundant.
Speaking of PalmOS, yes, it's a pity. The problem that the porter decided to redo whole backend and make it much more faster, he is just didn't came up to deadline, and he warned about it long beforehand. I'd suggest you to read scummvm-devel archives for details. So if we'd even postpone release because of PalmOS by 2 weeks, that wouldn't guarantee us presence of the port in final 0.8.0. All of us are volunteers, PalmOS porter included, and he just can't jump over his head.
Eugene
Thanks for taking the time to reply to my concerns. I do realize that this project is a completely volunteer based. I also find it to be one of the most professional volunteer based projects I have ever seen.
However, I still believe that 0.8.0 was rushed out the door. Using the Mac OS X build as example you stated that there are three areas where the Mac codebase differs the standard codebase. They are endianness, AltiVec and audio drivers.
You already mentioned that sword2 alignment problem (endianness) so, that is one out of three having an issue. Looking at the bug tracker shows priority 6 bug
[ 1342732 ] MAC: hq2x/hq3x filters broken when altivec is enabled in which fingolfin had to disable the AltiVec code in 0.8.0 in order for the hq filters to work. That's two out of three having an issue. Further up the bug tracker shows two possibly related priority 5 issues (correct me if I'm wrong)
[ 1344631 ] build fails on Mac OS X 10.2.8 when MT32 emulation is used and
[ 1343971 ] MAC: ScummVM 0.8.0 fails to launch which might be related to a conflict between Mac OS X 10.2.8 and MT32 emulation through the audio drivers. No one is assigned to either of these cases so, there is no telling if they are duplicatable or related or if they actually is related to the audio drivers so, this one is just a potential at this point.
So definitively two of the three Mac specific areas have issues in the 0.8.0 release. Depending on the outcome 1344631 and 1343971 there may be an issue in the audio drivers as well making it three for three. Had there been at least one official CVS build for Mac OS X after the Oct 11 call for testers I truly believe that these would have been caught and corrected before 0.8.0 release. So, in my opinion an official CVS build for Mac OS X is not redundant.
Thanks for clarifying the hold up on the PalmOS port; it would have been nice to have been informed in the main page ahead time that the PalmOS port wasn't going to be included in the 0.8.0 release to temper our disappointment. Since my programming claim to fame is 10 PRINT "Hello World", I've never felt that I was worthy of reading to the scumm-devel archives
![Wink ;)](./images/smilies/icon_wink.gif)