Error bei rootfs....

    Soooo, nächster error:



    arch.conf existiert, ist aber ein file. mkfs.jffs2 ist natürlich auch vorhanden...

    I briefly checked the source of mkfs.jffs2 to see where this error message was produced. Only place I could see was after a call to "scandir". The manual page for scandir suggests that it returns an error if it runs out of memory.


    So maybe you should check to see if you have enough swap enabled on the build machine?

    Thanks.


    It's a strange one. I can run the exact same commands by hand, and then it works. Just have to keep looking, I guess.

    I have the same problem here, on gentoo.
    Not always, but once every 3 builds or so.


    And the strange thing is that not only mkfs.jffs2 suffers from this, but also mksquashfs.
    The difference only is that mksquashfs doesn't abort with an error, it just logs a warning, and I end up with a squashfs image with arch.conf missing.


    I've tried to blame this to my parallel build options, but no, do_rootfs (which creates arch.conf) and image_cmd (which calls mkfs.jffs2 or mksquashfs) are correctly started after eachother, by the same process. Adding a 'ls -la' between the two even shows arch.conf is there, and it has the right attributes, and the right size.


    Sometimes arch.conf goes fine, but /usr/lib/ipkg/alternatives/[ (possibly the next inode) fails:

    Code
    mkfs.jffs2: opening directory /home/pieter/oe/pli-oe/build/tmp/rootfs/usr/lib/ipkg/alternatives/[: Not a directory


    Running mkfs.jffs2 (or mksquashfs) by hand with the same arguments, on the same rootfs, always succeeds by the way.
    Strange, and very hard to debug.


    If anybody has a bright idea, please let me know :winking_face:

    Zitat

    Originally posted by pieterg
    Adding a 'ls -la' between the two even shows arch.conf is there, and it has the right attributes, and the right size.


    Well, just decided to doublecheck this again (last time I looked into this was several months ago), turns out mkfs.jffs2 is right, at that time arch.conf has the 'directory' type, according to a 'ls -la' just after arch.conf is created.


    From build/tmp/work/dreambox-image-1.0-r0/temp/log.do_rootfs:

    Code
    + grep '^arch' /home/pieter/oe/pli-oe/build/tmp/work/dreambox-image-1.0-r0/temp/ipkg.conf
+ echo '------------------------------arch.conf created'
------------------------------arch.conf created
+ ls -la /home/pieter/oe/pli-oe/build/tmp/rootfs/etc/ipkg/
total 12
drwxr-xr-x  2 root root 4096 Mar 11 18:12 .
drwxr-xr-x 27 root root 4096 Mar 11 18:12 ..
drwxr-x---  1 root root   41 Mar 11 18:12 arch.conf


    When the build has stopped with the usual error however, arch.conf has changed into a normal file:


    Code
    ls -la tmp/rootfs/etc/ipkg/
total 16
drwxr-xr-x  2 pieter pieter 4096 Mar 11 18:12 .
drwxr-xr-x 27 pieter pieter 4096 Mar 11 18:12 ..
-rw-r--r--  1 pieter pieter   41 Mar 11 18:12 arch.conf
-rw-r--r--  1 pieter pieter   72 Mar 11 18:12 official-feed.conf

    Einmal editiert, zuletzt von pieterg ()

    Hello everybody,


    Got the same error here and this is what i did,


    From the point where my install stopt i edit the run.do_rootfs.1xxxx and removed the '-e' in the bbimage line.


    for type in jffs2; do
    if test -z "$FAKEROOTKEY"; then
    fakeroot -i /home/mythtv/OE_dm7020/build/tmp/fakedb.imag
    e bbimage -t $type /home/mythtv/OE_dm7020/openembedded/packages/images/dreambox-
    image.bb
    else
    bbimage -n "dreambox-image-dm7020-20080428185026" -t "$t
    ype" "/home/mythtv/OE_dm7020/openembedded/packages/images/dreambox-image.bb"
    fi
    done


    Quick and durty but i got my image and it is working perfect.
    O, and i did it this way simply because i can't find the script that is parsing the settings to the build files. :smiling_face:


    Maybe any one can explain this......


    Patrick

    I don't think removing -e actually does the trick. When I restart the build a few more times, eventually it will work, without any changes to the build command.


    Still haven't found the time to debug why this (apparent) race occurs.

    I have the same problem with my distri and my do_rootfs


    Code
    mkfs.jffs2 -x lzo --root=/usr/src/trunk/projects/foo/build/tmp/rootfs/../rootfs_g/ --faketime --output=/usr/src/trunk/projects/foo/build/tmp/deploy/images/rootfs.jffs2 --little-endian -e 128 --pad=27131904 
mkfs.jffs2: opening directory /usr/src/trunk/projects/foo/build/tmp/rootfs_g/lib/libblaconfig.so.0: Not a directory


    Sometimes i got an error on "cp".
    I have updated to ubuntu 8.04. The same error was on debian testing.
    Maybe this is a pyton problem


    Python 2.4.4 -> works
    Python 2.5.2 -> failes

    Hmmm, musste mein Kubuntu 8.04 auf meinem Rechner neu installieren und seit dem bekomme ich kein Image mehr gebacken:

    Zitat

    Originally posted by Wilko
    Same here kan anyone help please ?? :loudly_crying_face:


    for the original error? Update fakeroot and your problems are gone.
    For AliAbdul's error? I don't see how IMAGE_ROOTFS_SIZE_ext3 does anything, as we're not building an ext3 image. But it helped for him, so it might work for you as well.