Re: bin/59093: compile errors for userland on amd64 with gcc and -march=native

["Taylor R Campbell via gnats" <gnats-admin%NetBSD.org@localhost>]

The following reply was made to PR bin/59093; it has been noted by GNATS.

From: Taylor R Campbell <riastradh%NetBSD.org@localhost>
To: Onno van der Linden <o.vd.linden%quicknet.nl@localhost>
Cc: gnats-bugs%NetBSD.org@localhost, netbsd-bugs%NetBSD.org@localhost,
	christos%NetBSD.org@localhost, mrg%NetBSD.org@localhost
Subject: Re: bin/59093: compile errors for userland on amd64 with gcc and -march=native
Date: Wed, 26 Feb 2025 13:40:03 +0000

 I don't understand, why are any changes to the gcc *intrin.h files
 needed at all?  Shouldn't the warnings be suppressed under the default
 of -Wno-system-headers, unless you go out of your way to enable
 -Wsystem-headers?
 
 Initializing nadj in lgammal_r is harmless but we should mark it with
 
 	/* XXX gcc-12.whatever -march=native, PR bin/59093 */
 
 because the variable is not actually used uninitialized -- the
 initialization and the use are both conditional on `if (hx & 0x8000)'
 (and only on that).