The TsAGE engine for Ringworld now has sound playing! I'd be hard to pressed to express just how happy I was when I ran ScummVM and finally had correct sound playing back. I was like, Oh My God! it's finally working. After all the time spent re-implementing all the sound code from the original, and then spending nearly as much time slowly tracing through the execution of the original versus ScummVM to try and figuring out the cause of differences, it was a relief to finally have it working.
As such, the Ringworld game is that much closer to being complete. There are still a few areas left to look into, and polish up, such as:
* Sound fading isn't yet working correctly
* I need to hook the sound manager into the savegame format, and ensure playing sounds are persisted
* There are certain suspect areas in the sound manager's main _sfRethinkVoiceTypes method to review; at least one variable is being assigned to at one point and not currently being used, so I need to double check if that's really the case.
* The sound manager could use an overall clean-up.
It won't be too much longer until it's ready for public testing. Soon Larry Niven fans will be able to explore the Ringworld, and search for a way to defeat the bloodthirsty Kzins.
So what's coming next? Well, I've promised myself some time to work on Rex Nebular again, which I've somewhat neglected in my zest to have Ringworld completed. Strangerke is also enthusiastic about working on the other games that used the TsAGE engine, so I'll probably be spending some time on the game Blue Ice to determine what changes have been made to the engine, and provide a basis for being able to re-implement the game scene logic similar to what was done for Ringworld. Finally, I may offer to help out somewhat on the other games the team has received original source for. Gotta keep busy after all. :)
Tuesday, June 21, 2011
Monday, June 13, 2011
TsAGE sound support progressing
Work on the TsAGE sound has been progressing. I had originally hoped that the raw format used by the game would be a standard Adlib raw format that I could feed into one of the existing Adlib Sound FX classes. I tried to pass the data to the player class in the Cruise engine, which I had previously worked on, but unfortunately, whilst it didn't crash, it didn't play anything. And I'm not familiar enough with sound formats to have been able to diagnose why.
As such, I've spent some time powering through implementing all the relevant sound code directly from the demo executable. With the symbols identified, I at least had a heads up for the original engine names for a lot of the methods. Plus, we were able to garner some additional information from a demo version of the game Protostar. Whilst it's not an adventure game, the sound code has a lot in common with Ringworld; and not only did it have method names, it also had a lot of field information for the classes and structures. As with the Ringworld demo, it didn't have it for all the classes, but I was at least able to name fields in the sound classes and the sound manager.
From this basis I've now implemented what I think are all the relevant methods of the demo for both the Adlib sound driver and the sound manager classes - I'm currently using the OPL functionality with OPLWriteReg in place of direct port access. The resulting code I've completed doesn't actually play anything back yet, though. Making a log of the underlying writes to the Adlib port versus my ScummVM code, the results are different, so there's obviously some incorrectly implemented code.
My current major suspect is a method called _sfRethinkVoiceTypes - it's a enormously big method with over 500 lines, responsible for updating the voice channels based on queued sounds. It's one big mess, with a large pre-processing, post-processing, and main processing areas. I'll have to slowly trace through the execution both in ScummVM and the demo in the DOSBox Debugger to identify the problem. Hopefully once I've identified the differences, I'll finally hear some sound output.
Fingers crossed. :)
As such, I've spent some time powering through implementing all the relevant sound code directly from the demo executable. With the symbols identified, I at least had a heads up for the original engine names for a lot of the methods. Plus, we were able to garner some additional information from a demo version of the game Protostar. Whilst it's not an adventure game, the sound code has a lot in common with Ringworld; and not only did it have method names, it also had a lot of field information for the classes and structures. As with the Ringworld demo, it didn't have it for all the classes, but I was at least able to name fields in the sound classes and the sound manager.
From this basis I've now implemented what I think are all the relevant methods of the demo for both the Adlib sound driver and the sound manager classes - I'm currently using the OPL functionality with OPLWriteReg in place of direct port access. The resulting code I've completed doesn't actually play anything back yet, though. Making a log of the underlying writes to the Adlib port versus my ScummVM code, the results are different, so there's obviously some incorrectly implemented code.
My current major suspect is a method called _sfRethinkVoiceTypes - it's a enormously big method with over 500 lines, responsible for updating the voice channels based on queued sounds. It's one big mess, with a large pre-processing, post-processing, and main processing areas. I'll have to slowly trace through the execution both in ScummVM and the demo in the DOSBox Debugger to identify the problem. Hopefully once I've identified the differences, I'll finally hear some sound output.
Fingers crossed. :)
Subscribe to:
Posts (Atom)