Hi Folks,
Did anyone post some convenient starter MMC images for Debian on the XM
yet? (not Ubuntu). I like Ubuntu, but I would prefer Debian. I can of
course put it together myself, but that just takes a bit longer.
Jon.
I will once squeeze is released, but till then, use:
http://elinux.org/BeagleBoardDebian#Debian_Squeeze
It'll warn/error about the security repo's late in the install, just
skip it to the next section, bootloader i believe..
Tested on my xM A last week..
Regards,
Hey thanks. I'm actually doing the steps by hand as the host at home I'm
working on is Fedora (for various reasons...). Can you tell me which
toolchain you use for building these bits?
Also, just one thing on security. You commented out an "rm -rf" that
looks risky in that script. If $DIR is not set for some reason,
and /deploy exists, there's a theoretical problem. Same thing with
blindly following symlinks into /tmp when writing files. Sure, this is
probably on a single-user desktop, but I'd create a mktemp there.
Jon.
Oh, and can I also suggest that the serial.cmd file be explicitly copied
to the boot partition as serial.cmd/serial.scr and then renamed?
Otherwise it might confuse some people when the boot files are moved
around in the Squeeze post install "tweaks".
Thanks! I'm very happy you did this. It took some fiddling to do it all
manually but I did get the .35 kernel installed tonight. I will be very
happy to test a .36 install with working network once you have images
ready (I saw them on your site, but without a initramfs and I assume
that the directory naming difference with "d" vs. "x" implies something
too about the status of those files).
Jon.
Hi Jon,
Hey thanks. I'm actually doing the steps by hand as the host at home I'm
working on is Fedora (for various reasons...). Can you tell me which
toolchain you use for building these bits?
For a toolchain, i usually use the native gcc direcly on these
boards.. But as that can be quite slow, on x86 i use angstrom's cross
compiler..
Also, just one thing on security. You commented out an "rm -rf" that
looks risky in that script. If $DIR is not set for some reason,
and /deploy exists, there's a theoretical problem. Same thing with
blindly following symlinks into /tmp when writing files. Sure, this is
probably on a single-user desktop, but I'd create a mktemp there.
mktemp makes a lot more sense, and i had some time this morning, so i
converted it over..
https://github.com/RobertCNelson/netinstall-omap/commit/974d3c710fa072a6b3e6072231a249320a8a3f87
Regards,
Hi Jon,
Oh, and can I also suggest that the serial.cmd file be explicitly copied
to the boot partition as serial.cmd/serial.scr and then renamed?
Otherwise it might confuse some people when the boot files are moved
around in the Squeeze post install "tweaks".
The only reason it's boot.cmd right now, as there is a script
"./tools/boot_scripts.sh" on the fatfs for rebuilding a tweaked
'boot.cmd' but i can tweak that script to look for both, then it could
remain as serial.cmd..
Thanks! I'm very happy you did this. It took some fiddling to do it all
manually but I did get the .35 kernel installed tonight. I will be very
happy to test a .36 install with working network once you have images
ready (I saw them on your site, but without a initramfs and I assume
that the directory naming difference with "d" vs. "x" implies something
too about the status of those files).
The initramfs takes an active root file system to generate, it's a
manual build and upload.. Hence there's only a couple uploaded on
rcn-ee.net
The 'd' is really a development image, it's tested on one of my
machines (but right now, it's hardlocking on bootup on my c4 board and
i'm not sure why...).. The 'x' or 'l' is closer to stable, it's boot
and stress tested on 2-3 different omap3 platforms running 24/7...
Regards,