Sorry for all the mailing list email spam (along with my internal team emailing habits). I'm sure you're all tired of hearing from me about this but I wanted to post anyway. I tried using a v7hl builder to make a v6hl buildroot with the pkgs I built. It seems to chroot inside of it ok but when I take a rasp pi sd card and replace the rootfs on it I get :: Kernel panic - not syncing: Attempted to kill init! I took a blurry screen shot of it here: http://i.imgur.com/aaQeq.jpg
Anyway, I have a bad feeling that I messed up the builds I've done for the rasp pi. I think it may have been because I didn't use the armv6hl glibc/gcc pkgs when building the other pkgs afterwards so maybe gcc put in the wrong op codes or machine code instructions/translations during compile time? I'm not sure at all how compilers work and whether or not v6 assembly can be generated with a v7 compiler? I'm a bit lost and I'm pretty much just guessing at this point but whatever, there's not much I can do I'm assuming, heh... I gave it my best shot and tried anyway :/ :S
As you may or may not know, I've been assigned the task of trying to bring up a semi-new arch for the rasp pi. I think this is a bit too much for me to handle as I am not a programmer and when things go wrong besides package build failures, I get stuck pretty easily. So my status is, I've been using a mixture of rpmbuild && mock on an armv7hl machine to compile around 750+ core packages and I've been using rpmrc && macro files which contain the following compiler flags:
... -march=armv6 -mfloat-abi=hard -mfpu=vfp ...
With my initial set of packages, I tried to install them on a armv5 rasp pi using this command:
rpm --force --ignorearch --nodeps --root /root/f17arm6hl/ -i ./repo/*
The first sign that something must have gone wrong was all the "%post" scriplet warnings that were being displayed during the rootdir install. Once rpm finished, I tried to chroot into that directory to see if it would run. The result is the following:
[root@raspi ~]# chroot /root/ f17arm6hl/
As you can see, this is where the problem suddenly goes way above my head. I don't know much about how to debug what's really going on here but it seems like I may have compiled a possible library or binary with a v7hl instruction? Not sure how to check for that though... Could it be that chroot'ing into a v6hl directory from a v5 rasp pi is preventing some sort of initial machine code symbol lookup to fail? If there's anyone who has any recommendations on how I could debug or backtrack this on a lower-level maybe I'd be really interested in hearing any suggestions you have. Maybe my whole approach in trying to load or bootstrap these initial packages is wrong, I'm not sure.
I did try to perform an objectdump of a sample v6hl binary (/bin/rpm) to see if there were any instruction/op code complaints and there did exist some lines which contained the following ( http://pastebin.com/WmK9tQJt ) :<http://pastebin.com/WmK9tQJt>
[root@raspi ~]# objdump -d /root/f17v6hl/bin/rpm
98c4: 00000afc strdeq r0, [r0], -ip
98c8: 000007b4 ; <UNDEFINED> instruction: 0x000007b4
98cc: 00000bdc ldrdeq r0, [r0], -ip
So I'm stuck and need some advice. If you need any other files posted or commands run to help try and debug this let me know!
Thanks for your time,
Hello, ARM SIG. My name is Russell Golden. I package
mozilla-https-everywhere for Fedora and EPEL, and co-maintain
mozilla-adblockplus for the same.
I'm going to add Fedora to my r.fp.o side repo for
mozilla-https-everywhere. I was wondering if you would like to see ARM
supported in it, and if so how you would prefer I set up the arch.
It's a noarch package, so I shouldn't have any problems making it available.