Well, we have a problem with TraceMonkey already. This is important -- if you are a 601-based Classilla user, I need to hear from you ASAP (post in this thread).
TraceMonkey fetches the processor's timebase using mftb/mftbu instructions. While these technically exist on the 601, they work completely differently from the 603 on up, and CodeWarrior refuses to use them or compile (if I force the issue with 'machine 601' I get a compilation error). This is a critical portion of the nanojit's inline assembly code and cannot be coded around. It does compile if I use 'machine 603' or 'machine 604', so it is clearly an architectural limitation.
This means that 6100, 7100, 8100, 7200 and 7500 users, and any other 601 Power Macintosh, cannot use TraceMonkey as written. I need to know *now* if you are such a user -- if there is a significant number of 601-based systems out there using Classilla, I will not continue with TraceMonkey at this time. I can't even guarantee I can port it to the 603 or 604, although if I encounter incompatibilities with those, then TraceMonkey is a no-go because I know of many people using systems with those CPUs. However, I already know that the 601 is dead in the water unless there is someone with bigger PPC assembler gonads than I that knows a solution to this. (johnklos? trag?)
This should not affect you if your system has an upgrade card in it -- I am pretty confident that the instruction would be handled by the upgrade card, not the internal CPU, even on systems where the original internal CPU remains present. I can't promise this, however, especially for weirder upgrades like the 7200 Sonnet CPU card.
If I am unable to port TraceMonkey, and it was sort of a long shot anyway, the updated SpiderMonkey will still be in Classilla 9.2 and even that is still a significant improvement, so this is hardly catastrophic. But please, speak now if you are using a 601 system with Classilla so I can at least do some future planning.