[oe] OE beginner constantly failing to build anything

Koen Kooi koen at dominion.thruhere.net
Wed Oct 12 20:39:05 UTC 2011


Op 12 okt. 2011, om 22:34 heeft Gary Thomas het volgende geschreven:

> On 2011-10-12 13:08, Koen Kooi wrote:
>> -----BEGIN PGP SIGNED MESSAGE-----
>> Hash: SHA1
>> 
>> Op 12-10-11 20:25, Khem Raj schreef:
>>> On 10/12/2011 5:22 AM, Rainer Koenig wrote:
>>>> Hi there,
>>>> 
>>>> the following lines might give you the feeling of a rant, sorry about
>>>> that but after 2 weeks of no success I need to write down quickly what
>>>> bothers me. Please don't feel offended by my deep desperation. ;-)
>>>> 
>>>> My build system is running Debian/Squeeze (amd64) and I also
>>>> downloaded the CodeSourcery Toolchain. Was able to compile a simple
>>>> hello world with that for Angstrom, but now I wanted to build Angstrom
>>>> from scratch.
>>>> 
>>>> So I started with OE classic, tried the "getting started" from the PDF
>>>> docs and failed. I learned quickly that there are version dependencies
>>>> between bitbake and the version that I checkout of the OE tree. I also
>>>> learned that some of the failures come because I tried to bild "old"
>>>> images and meanwhile the sources that bitbake wants to fetch seem to
>>>> have moved to another place or got deleted. I tried according to the
>>>> wiki docs that said OE 2011.3 works with bitbake 1.12.0 but even then
>>>> I failed. Last failure had to do with building libtool-2.4-cross when
>>>> the build system told my that my compiler can't create binaries...
>>>> after the process has build around 60+ packets without any problem.
>>>> 
>>>> Well, I learned that OE classic is no longer maintained and that I
>>>> should switch to OE-core. So I said, lets start from scratch again and
>>>> see. I followed the Angstrome building instructions as they are
>>>> written here:
>>>> http://git.angstrom-distribution.org/cgi-bin/cgit.cgi/angstrom-layers/tree/README
>>>> 
>>>> 
>>>> 
>>>> 
>> Did everything that was described there and faild again. This time
>>>> with the following errors:
>>>> ----------------8<-snip---------------------- bitbake base-image Pseudo
>>>> is not present but is required, building this first before the main
>>>> build NOTE: angstrom DOES NOT support libiconv because the eglibc
>>>> provided iconv library is used                         | ETA:
>>>> 00:04:17 NOTE: angstrom DOES NOT support libiconv because the eglibc
>>>> provided iconv library is used                         | ETA:
>>>> 00:04:16 NOTE: angstrom DOES NOT support gconf-dbus because gconf-dbus
>>>> has been merged back into main GConf                 | ETA:  00:03:57
>>>> NOTE: angstrom DOES NOT support gconf-dbus-native because gconf-dbus
>>>> has been merged back into main GConf ERROR: Error Method already seen:
>>>> get_tls_setting in' glx-use-tls.inc' now in 'glx-use-tls.bbclass'
>>>> | ETA:  00:02:59 ERROR: Could not include required file
>>>> recipes-graphics/xorg-xserver/xserver-xf86-dri-lite.inc | ETA:
>>>> 00:02:01 ERROR: Command execution failed: Exited with 1
>>> 
>>> bad timing. There is a bit of churn due to xorg updates happening in
>>> core layer and other layers need to adjust once the core parts are
>>> settled in. This is master so expect few bleeding here and there
>>> occasionally
>>> 
>>>> ----------------8<-snip----------------------
>>>> 
>>>> Then I read "glx-use-tls" and remembered to saw that in the list
>>>> archive of today. There is a mail from Khem Raj Subject: [oe]
>>>> [meta-oe][PATCH 1/2] classes/glx-use-tls.bbclass: Delete
>>>> 
>>>> Yeah. I checked my meta-oe tree and glx-use-tls.bbclass is still
>>>> there. Then I checked where that tree came from, it comes from github,
>>>> but nevertheless, also the tree from http://git.openembedded.org at
>>>> the moment has no patch like that applied.
>>>> 
>>>> So I need to start questions:
>>>> 
>>>> 1. Is there any *failsafe* procedure that lets me build some image for
>>>> my beagleboard? I'm relatively new to embedded linux and I need to
>>>> read a lot about bitbake and oe, but failing with every "getting
>>>> started" instructions doesn't give me much confidence that I will ever
>>>> succeed.
>>>> 
>>>> 2. How are patches applied in OE? The patch from Khem Raj is not yet
>>>> in the git-tree, so how can I patch it to my tree.
>>> 
>>> with some git-fu you can. (git am) can help
>>> 
>>> However those patches are just a workaround to keep parsing happy they
>>> are not final patches that should go into repo.
>>> 
>>>> 
>>>> 3. I tried to search for the bugs I encountered and found out that you
>>>> have abandoned a tool like Bugzilla in favor of reporting bugs on the
>>>> mailing list. From my perspective the mailing list is not very helpful
>>>> to *find* information about existing bugs and their status, so is
>>>> there a reason for this switch that I can understand?
>>> 
>>> It has been discussed extensively and decision to have mailing list was
>>> the outcome. You can searc through mailing list archives to gmane will
>>> help you.
>>> 
>>>> 
>>>> I think there are solutions for my problem, the fact that Angstrom
>>>> exists and can be downloaded makes me think that somebody was able to
>>>> build it from the sources, but it seems that I got the wrong moment in
>>>> time for my attempts so all the sources and recipes that I'm depending
>>>> on are constantly changing and I always get lost in the gory details.
>>>> 
>>> 
>>> generally everybody builds from source every day its a norm with
>>> openembedded.
>>> 
>>> You could try the 2011.03 release branch which should be pretty stable
>> 
>> Especially if you're using the angstrom setup script instead of some 'pdf'
> 
> Sadly, even this magic script is often broken.
> 
> I just updated the script to rev
> I then ran (from scratch):
>  % MACHINE="beagleboard" ./oebb.sh config beagleboard
>    ... lots of update messages, but these errors:
>  layer repository name: meta-openembedded
>  layer uri: git://github.com/openembedded/meta-oe.git
>  layer branch/revision: master/HEAD
>  Updating layer at layers/meta-openembedded
>  fatal: read error: Connection reset by peer

Yes, connection resets are indeed the fault of the script. As is the economic crisis and electricity outtages.



More information about the Openembedded-devel mailing list