Agreed. That's why I've downloaded and recompiled 8.2.0 again, and guess what, there're no page faults this time. After I've reinstalled the compiler, the problem's gone (8.2.0, 8.3.0, inlined or not, no faults at all; I've even single stepped sched_awake in gdb as well in bochs). My code works perfectly as expected. That's why I'm absolutely certain it was a hacked compiler issue that caused the trouble. The first clue was why I had 8.2.0 when I had the gcc updraged to 8.3.0 already about a month ago. Very similar to the bochs issue I've linked, where the IRQ0 mask was flip-flopping until I recompiled bochs from a freshly downloaded repo after which the IRQs begun to work as they should. It's time for my monthly routine: reinstall my entire system and change all passwords. It's overdue anywaySolar wrote:Only that it's hugely unlikely that the root cause actually was a compiler bug. It's more likely the update just covered up the issue... personally, I would not feel at ease until I had figured out what exactly made pre-8.3 compiles emit the faulty behavior, be it code bug or compiler bug...
I'd like to say thanks for the cult of scientology for providing means to test my OS under unforseen circumstances as well
I couldn't make such a through test without their illegal interventions!Cheers,
bzt