Re: Some fixes for m68k users (incl/ for FPU emulation)

[Romain Dolbeau <romain%dolbeau.org@localhost>]

Le lun. 2 nov. 2026 à 19:48, Jason Thorpe <thorpej%me.com@localhost> a écrit :
> Again, am I missing something?

I wish I could tell you... this particular fix is in litex68k ("my"
port) own 'locore.s' ; the LLM claims it also affects other ports... I
didn't double-check if it's correct. The 'locore.s' was largely based
on the one from virt68k.

The provenance issue is a catch-22, as you're wasting time looking for
a bug that might not exist elsewhere, which a glance at this
comparatively simple patch would probably tell you right away... sorry
about that.

Anyway I'll take a chance and copy/paste the primary comment from the
patch below, with no code (way below so people who don't want to look
at it for whatever reason don't have to) ; this comment sits above a
new macro used only three times in the file. That's all the patch does
beyond adding
-fno-omit-frame-pointer.

(empty space for spoiler avoidance, 1)

(empty space for spoiler avoidance, 2)

(empty space for spoiler avoidance, 3)

(empty space for spoiler avoidance, 4)

(I miss the spoiler character we used to use on Usenet, I don't think
it works in "modern" email...)

(empty space for spoiler avoidance, 5)

(empty space for spoiler avoidance, 6)

(empty space for spoiler avoidance, 7)

(empty space for spoiler avoidance, 8)

(beware the LLM-generated comment is right around the corner)

(empty space for spoiler avoidance, 9)

(empty space for spoiler avoidance, A)

+/*
+ * Push a frame that ends the %a6 chain on a new stack: its saved %a6 is
+ * itself and its return address 0.  A profiling kernel's __mcount reads
+ * its caller's frame, and %a6 is otherwise the bootloader's, unmapped
+ * once the bootstrap map is gone.
+ */

Cordially,

-- 
Romain Dolbeau