On Tue, Oct 17, 2023 at 02:38:40PM -0500, Justin wrote: > Okay, so because the VIA C3 Nehemaiah chip doesn't properly implement > ENDBR32, it falls outside of the supported hardware, despite being otherwise > an "i686 class chip," correct? Seems correct to me.
> Or put another way, in classic Cyrix style, it's "almost a 686." > > As for what I was looking for, well, two things: > > 1. Clarification on the "big picture" situation in the most precise terms. > Assuming the answer to my question above is "yes," then this part is covered. Your GCC link, assuming it's the current code, suggests that GCC thinks "supports CMOV means supports ENDBR32". Nehemiah does the former but not the latter. And binaries compiled with -march=i686 require CMOV support so they require ENDBR32 support. In other words, yes, VIA C3 Nehemiah doesn't seem to be a supported CPU for -march=i686 (even if that worked before). > 2. If this was a bug, I was going to ask if there was anything else I > could do to help fix it. Since it looks like it's not a bug - not in > the SOFTWARE, anyway... - is the best suggestion for hardware like this > to use Debian LTS (and bullseye) for the foreseeable future? Yes, assuming the pre-bookworm Debian i386 architecture fully supports it, as I don't know what *exactly* was allowed in the "almost i686" stretch-bullseye i386.
