On Fri Jan 24 19:06:36 2025 +0000, William Horvath wrote:
It's not "so difficult", I closed it because the MR doesn't make sense to me anymore if it's no longer just focused on fixing what's broken. This approach generally feels wrong to me if, to continue with fixing the PE-side, we'd have to do something similar for dozens of other occurrences. Like I mentioned, I believe there is a better ("lazier") way to make sure the compiler behaves properly with inline asm function calls, instead of manual additions. I will explore other options.
There's no better way. You simply have to let the compiler know. And like I said, you don't have to do the others if you don't want to, in fact I very much prefer to keep this MR short and only focused on minimal amount of components (or unix side as you did), so we can get the macro and "infrastructure" in first.
For your concern about doing it for everything: there's not that many ASM_GLOBAL_FUNCs that call other functions around in general, it's not like the PE conversion that touched every single module. But I could even take up that task if you want, as long as we get this one in first so I know how to proceed.