I've been trying to reproduce the bug on the hardware available to me (5 different PCs), but I can't (yes, I unplugged the PS/2 mouse).
Well, I have only one machine to test ReactOS on real hardware. And I also don't see the problem.
So the following statement ("...most real hardware") in Bug 688 is very questionable:
BZ> The addition of i8042prt created some changes in console allocation during BZ> text-mode setup, which leads to AllocConsole failing (for unknown reasons BZ> there). This stops installation on most real hardware.
If we can't fix it, maybe we should put the pre-i8042prt code back and use that for now? Would that cause major other problems (e.g. PnP issues)?
There is a comment in the related Bug 682, which may be a helpful idea:
BZ> When I started ros install (boot cd) I switched my keyboard and monitor to other BZ> computer and when setup started to scan for devices it didn't find keyboard and BZ> gave a blue screen. So maybe we should implement some kind of fallback BZ> system so that if for example keyboard isn't found it will fallback to a basic BZ> keyboard and use that.
If I understand correctly the problem lies in the mouse detection code. On some hardware the detection goes wrong. So let's at least disable the mouse, but avoid to stop the setup. Using the keyboard without the mouse should be enough in this stage.
Regards,
Martin