Re: unexpected riscv64 build success on buildd / NaN payload

[Rene Engelhard <rene@debian.org>]

Hi,

Am 27.02.24 um 19:26 schrieb Rene Engelhard:
> Then I more wonder why it failed the test on -03 and "worked" on -05 now 
> where the difference is just -O0 vs. -O2.

I did "quick" tests (with some "shortcut" hacks to get the test ran 
without building everything).

Anyways:

Indeed:

- milkv: O2 "works", O0 fails
- ricci: O2 "works", O0 fails

I could swear in my tests before I did enable O2 in the experimental 
upload it also failed there.

I guess gcc did change....

Reverting that change (O0 is closer to Os which is in the upstream 
makefile anyway) for now to be sure. I am totally not sure that keeping
O2 here and thus having the test pass is a good idea, especially since
if the hardware does not have that feature. Does it?

@admin, @wb-team: nevermind
Rest should probably be done in -riscv.

Regards,

Rene