Re: Why does this have to be so ##@&$@$! difficult?

Sorry, I did try to give enough discussion in my last answer to illustrate the general approach, as well as the specific.  To be completely general then:

a) check for the version you need in the chumby-oe tree.  It
may be there, but not being pulled unless you explicitly change the version
in your image recipe.

IF NOT

b) check in the head of the openembedded repo and if it is there, copy it over along
with any support files.  You may have to move more packages to resolve dependencies.

In an ideal world our OE tree would regularly merge in changes to the head, but the are changes in the head that
make the Chumby unbootable, and AFAIK no one has ever identified which
updates or why.

IF NOT

c) Search around for someone else's work that isn't in the openembedded master tree.
There are lots of people building OE recipes and posting them to various places.

FINALLY

d) Build it yourself.  Start with the bb file you have, copy it,
start editing and testing.  Sometimes this turns out to be trivial, sometimes very much non-trivial.

Yeah, it's hard.  When we eventually get the 3.4 kernel and post-classic(?) OE booting on the Chumby this should be a lot easier because we will be operating with the current kernel and OE sources.   But there is a lot of work to do to get there.

Re: Why does this have to be so ##@&$@$! difficult?

This recipe file for nodejs 0.6.17 builds cleanly.  The crucial addition is this line:

  export CCFLAGS="-march=armv5te"

I've started a clean build now and will do a smoke test once it completes.

Re: Why does this have to be so ##@&$@$! difficult?

Dang it!  I keep geting this error:

NOTE: package chumby-falconwing-bootimage-1.0-r8: task do_deploy: Started
ERROR: Logfile of failure stored in: /home/chris/chumby-oe/output-angstrom-.9/work/armv5te-angstrom-linux-gnueabi/chumby-falconwing-bootimage-1.0-r8/temp/log.do_deploy.30263
Log data follows:
| + do_deploy
| + config_util --cmd=create --mbr=/dev/zero --configname=falconwing --build_ver=1000 --force --pad --blockdef=/home/chris/chumby-oe/output-angstrom-.9/deploy/eglibc/images/chumby-falconwing/chumby_shell.bin,215040,boot,1,0,0,0 --blockdef=/home/chris/chumby-oe/output-angstrom-.9/deploy/eglibc/images/chumby-falconwing/4_1.bin,153600,img1,1,0,0,0 --blockdef=/home/chris/chumby-oe/output-angstrom-.9/deploy/eglibc/images/chumby-falconwing/4_2.bin,153600,img2,1,0,0,0 --blockdef=/home/chris/chumby-oe/output-angstrom-.9/deploy/eglibc/images/chumby-falconwing/5_1.bin,153600,img3,1,0,0,0 --blockdef=/home/chris/chumby-oe/output-angstrom-.9/deploy/eglibc/images/chumby-falconwing/5_2.bin,153600,img4,1,0,0,0 --blockdef=/home/chris/chumby-oe/output-angstrom-.9/deploy/eglibc/images/chumby-falconwing/5_3.bin,153600,img5,1,0,0,0 --blockdef=/home/chris/chumby-oe/output-angstrom-.9/deploy/eglibc/images/chumby-falconwing/zImage-chumby-falconwing.bin,4194304,krnA,1,0,0,0 --blockdef=/home/chris/chumby-oe/output-angstrom-.9/deploy/eglibc/images/chumby-falconwing/zImage-chumby-falconwing.bin,4194304,krnB,1,0,0,0
| config_util v0.23
| Cannot open /home/chris/chumby-oe/output-angstrom-.9/deploy/eglibc/images/chumby-falconwing/zImage-chumby-falconwing.bin for validation: errno=2 (No such file or directory)
NOTE: package chumby-falconwing-bootimage-1.0-r8: task do_deploy: Failed

What is up with that??!!

Re: Why does this have to be so ##@&$@$! difficult?

I must be learning!  I ran make unit (on a hunch) and built my image.

An old dog CAN be taught new tricks!

Re: Why does this have to be so ##@&$@$! difficult?

I spoke too soon.  I can't get a silver-moon image.  I'll try make really clean and see what shakes out.

Re: Why does this have to be so ##@&$@$! difficult?

Hey GuyC

I was just over here:

https://github.com/clearwater/chumby-oe … 18e#diff-1

and noticed the lines :

-export MACHINE:=chumby-falconwing
     14   
+export MACHINE:=chumby-silverwing

Shouldn't that be silvermoon?

Re: Why does this have to be so ##@&$@$! difficult?

Quite so, but you will notice there was a subsequent commit with the commit message "Fix makefile typo - derp derp derp" which rectifies that smile

Re: Why does this have to be so ##@&$@$! difficult?

BTW, I believe my recipe for nodejs 0.6 is wrong.  I suspect that the change to CCFLAGS may be affecting other packages.  First I will confirm that is the cause of my problems and then I'll figure out how to fix it.

For the record the failure I'm seeing looks like this:

| config_util v0.23
| Cannot open /home/guy/chumby-oe/output-angstrom-.9/deploy/eglibc/images/chumby-falconwing/zImage-chumby-falconwing.bin for validation: errno=2 (No such file or directory)
NOTE: package chumby-falconwing-bootimage-1.0-r8: task do_deploy: Failed

zImage-chumby-falconwing.bin is not being created, but there are no previous errors in the build.

Re: Why does this have to be so ##@&$@$! difficult?

I've been having that problem for days now.  I thought it was something I did wrong.  I've deleted chumby-oe twice and rebuilt the whole thing.  Right now I'm building after "make reallyclean".

Re: Why does this have to be so ##@&$@$! difficult?

It appears to me that the build error we are both seeing is caused by commenting out

INHERIT += "rm_work"

from meta-chumby/conf/local.conf.  I've been rolling back changes and only after undoing that change was I able to get a clean build.

I think it shouldn't be that way, but there you go.

With that cleared up I will return to the nodejs 0.6 issue!

Re: Why does this have to be so ##@&$@$! difficult?

The good news is that I can confirm that after putting rm_work back in, I can get nodejs 0.6.17 to compile and install on the CHB.

root@chumby-falconwing:~# which node
/usr/bin/node
root@chumby-falconwing:~# node --version
v0.6.17

The bad news is that it doesn't actually run.  Running it against a trivial or even empty JS file gives this error:

Illegal instruction

I guess V8 isn't correctly configured for the platform.  I see others with similar looking problems running nodejs on arm platforms.  This thread seems promising, and references a page on CrossCompilingForARM.

For completeness here's my current version of nodejs_0.6.17.bb

DESCRIPTION = "nodeJS Evented I/O for V8 JavaScript"
HOMEPAGE = "http://nodejs.org"
LICENSE = "MIT"
DEPENDS = "openssl"

SRC_URI = " \
  http://nodejs.org/dist/v${PV}/node-v${PV}.tar.gz \
"

SRC_URI[md5sum] = "55641d5d280cdd4e0c030c203107599d"
SRC_URI[sha256sum] = "8dfe5948de27e37a14af184f06e7bd89a23c3b248af44c8ef5cffcd0e4c65778"

S = "${WORKDIR}/node-v${PV}"

# v8 errors out if you have set CCACHE
CCACHE = ""

# v8 wont build on less than armv5t
do_configure () {
  ./configure --prefix=${prefix} --without-snapshot
}

do_compile () {
  oe_runmake CCFLAGS="-march=armv5te"
}

do_install () {
  DESTDIR=${D} oe_runmake install
}

BBCLASSEXTEND = "native"

Re: Why does this have to be so ##@&$@$! difficult?

guyc wrote:

It appears to me that the build error we are both seeing is caused by commenting out

INHERIT += "rm_work"

from meta-chumby/conf/local.conf.  I've been rolling back changes and only after undoing that change was I able to get a clean build.

I think it shouldn't be that way, but there you go.

With that cleared up I will return to the nodejs 0.6 issue!


I was wondering about that...

88 (edited by quispie12@yahoo.com 2013-09-30 16:19:27)

Re: Why does this have to be so ##@&$@$! difficult?

Hey folks,

I'm looking for an OE image for my Insignia 8, hopefully with working touchscreen support.

So far I've tried to recreate the original build environment on an Ubuntu 10.04 machine but with little luck.

I"ve been trying these 2 build howtos with varying results:

http://wiki.chumby.com/index.php?title= … ded_(Beta)    (created silvermoon image but it doesn't quite boot up.  it freezes after the startup screen due to interrupt errors)

http://www.kosagi.com/w/index.php?title … enEmbedded      (fails 12 hours into the build due to GIT dumb-https errors)

Can anyone provide assistance?