That means the loader could not allocate that much stack space to your process, so you get a null pointer. If this is the browser, it's likely because your code in memory is much bigger than TenFourFox's was (you get about 2GB total which must be split between your code pages, data pages and the...
If you don't do it, you'll get crashes in the browser trying to run any code of substance. The test suite will mostly run, but most of those are trivial. Because so many registers need to be saved in each stack frame and calls can be recursive, it's pretty much unavoidable.
Yes, it was a lot easier debugging it when I could run automated tests. The command line shell was massively faster at that. It won't catch every bug but it catches a lot.
Be mindful of the fact that the JIT has monstrous stack requirements. There is specific code in the TenFourFox build system...
I messed around a lot with CPU Director trying to make it happy. With speculative execution off (!) and all caches off (!!), it works sometimes (!!!). I have to find out what type of chicken it wants sacrificed to it.
Well, I think you win, outside of unusual lab setups.
I do have several radioSHARKs. Not PCI, but USB and they work with nearly everything, including Power Macs (my G4 server is configured to record shows from it and play its output to a house FM transmitter...
It's called an HDI-30 connector. There are many passive converter cables for it on eBay. Pick the BlueSCSI or ZuluSCSI that works with the cable you choose.
That would be fabulous. I've been spending the afternoon messing with Boot Variables' source code but it does not like New World Macs like my development MDD. I'm trying to adapt @joevt's NVRAM read code from DumpNameRegistry and it no longer crashes but there seems to be an impedance mismatch I...
Setting the property to 1 from Open Firmware indeed enables on-board Ethernet with 1.1.20.1.
Now to think about if we want to solve this with the nvramrc or some other means.
I've been looking at the Ethernet issue some more.
DumpNameRegistry shows two differences for /Devices/device-tree/bandit/gc/mace: 2.0 has a property PLSCCPort 00000001 which I doubt is relevant to the problem, and 1.1.20.1 has a different driver-ist InterruptSetMember offset. In the 2.0 ROM...
I have an 8*24 GC in my MacIvory system and took some benchmarks a while back.
https://oldvcr.blogspot.com/2024/10/refurb-weekend-symbolics-macivory-lisp.html
Make an image of the 8.1 CD as an iso, then set it up on an external BlueSCSI/ZuluSCSI or similar. They can directly emulate a CD-ROM from an .iso; it's all in their documentation.
Something just occurred to me. If you hook up the speaker leads to that card, do you get audio output from the machine through the line out? Maybe it's mixing the machine's built-in sound.