Donald Haase

> Is this because you found that the va_list processing is inherently using `malloc()` internally, and doing it like this gives you a form that is malloc()-free? Right, either the...

> Can't we use the existing ramdisk code for that? > > `/ram` could actually be moved to be the root (which would become R/W), allowing for directories to be...

I've found similar leaks in mprec now in serpent_dma as well. That one, when I was doing the malloc debug testing for #599 , did not have those memory leaks,...

Now @gyrovorbis whenever you could rebase this one I'll take a look :) It should be less than half the number of lines of changes and they should pretty much...

I was ready to merge and went back to retest the example and it seems that it just hangs for me now after: `Testing accepting the exception in GLOBAL handler......

Turns out my other system was on stable, so couldn't build the example for this. After rebuilding my whole environment to 14.2.0 I now get this:  And the example...

> TBH this PR makes me uneasy. The idea that until now all IRQs handlers were assumed as running in atomic context, and this PR changes that, sounds like opening...

> We kind of half-assedly already did support multiple layers of IRQs--we just simply immediately shit our pants when the second IRQ was raised and were never even able to...

@gyrovorbis - I did the backend work to clean up and rebase this in the branch https://github.com/KallistiOS/KallistiOS/tree/irq_cleanup_final . If it looks good to you, you could go ahead and just...

Did one more round to get the whitespace cleanups and the one extra error check into master out of that `irq_cleanup_final`. The branch is now just 3 commits with the...