Date: Mon, 23 May 2016 22:51:45 -0700 From: bob prohaska <fbsd@www.zefox.net> To: Warner Losh <imp@bsdimp.com> Cc: Ian Lepore <ian@freebsd.org>, "freebsd-arm@FreeBSD.org" <freebsd-arm@freebsd.org> Subject: Re: Confirmed update procedure for a self-hosted armv6hf system Message-ID: <20160524055145.GE52783@www.zefox.net> In-Reply-To: <CANCZdfqosAci0_CbbWa09_AMoWPHSLeEjN1%2BKqwRRq7wZF1sSg@mail.gmail.com> References: <1463976239.1180.406.camel@freebsd.org> <20160523215248.GA52783@www.zefox.net> <1464055880.1180.433.camel@freebsd.org> <20160524024027.GB52783@www.zefox.net> <1464058197.1180.439.camel@freebsd.org> <20160524031855.GC52783@www.zefox.net> <1464060955.1180.441.camel@freebsd.org> <20160524043635.GD52783@www.zefox.net> <CANCZdfqosAci0_CbbWa09_AMoWPHSLeEjN1%2BKqwRRq7wZF1sSg@mail.gmail.com>
next in thread | previous in thread | raw e-mail | index | archive | help
On Mon, May 23, 2016 at 10:44:17PM -0600, Warner Losh wrote: > > If I did the job right, the armv6hf binaries will be the same as the new > armv6 binaries (with one minor niggle in how the elf files are marked, > but that marking isn't used). > So if I start with an old installation that reports in part % uname -a FreeBSD www.zefox.net 11.0-CURRENT FreeBSD 11.0-CURRENT #2 r297367M: Tue Mar 29 20:38:59 UTC 2016 root@www.zefox.org:/usr/obj/usr/src/sys/RPI2 arm fbsd@www:~/public_html/rpi2/crashes/crash_5_21_16 % uname -p armv6 the upgrade won't "know any better" and simply work? An older kernel is available to try that. > All the naming mismatches are just in the name. Of the versions reported by uname, or something else, perhaps in a Makefile? > > I'm surprised that > > setenv MACHINE arm > setenv MACHINE_ARCH armv6 > setenv UNAME_p armv6 > make buildworld > make buildkernel > make installkernel > make installworld > > wouldn't work, even if the kernel reported armv6hf (the defines are important > but the order of the makes likely doesn't so long as buildfoo comes > before installfoo, > and rebooting with the new kernel or continuing to use the old one. > Buildworld and buildkernel both report successful completion. Is there any plausible way this could be untrue? Missing pieces would certainly cause the install to fail, though I'd think with "file not found" rather than suggestions to set a compiler type. The purpose of this system is to experiment, I'd rather not break it beyond repair but sometimes that's the price of progress.... Any ideas are welcome! bob prohaska
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?20160524055145.GE52783>