* tar & ar native vs host (Fedora 9 issues)
@ 2008-06-10 11:50 Terry Kemp
2008-06-11 1:07 ` Terry Kemp
0 siblings, 1 reply; 3+ messages in thread
From: Terry Kemp @ 2008-06-10 11:50 UTC (permalink / raw)
To: openembedded-devel
Hello devs,
I have been working on http://bugs.openembedded.net/show_bug.cgi?id=4350
and I have attached my recipe mods to this for review should someone
care to look at them.
To explain though...
Fedora 9 builds won't get past do_rootfs - bad stuff happens in
arfile.py.
First I tried forcing ipkg builds with tar-native. It didn't fix _this_
problem but I think it is a useful thing to do anyway. Dunno what you
will think of the brute force and ignorance approach in
ipkg-utils-native tho :-D
Next I have pretty much confirmed the Fedora 9 ar command doesn't play
nice with OE.. it inserts spurious ` characters into the ipkg.
To fix this I had to create a binutils-native recipe (its a bit ugly but
it gets me a native ar). Is there already a native staged ar somewhere?
The cross one dinnae wanna work with x86_64 anyway.
This seems to sort the issue (at least from looking at the ipkg in hex
editor) - although I am yet to confirm this, will know in about 4 hrs -
on about my 40th build from scratch.
5 hrs later... - IT WORKS!
NOTE: Running task 1296 of 1299 (ID:
4, /home/tmk/gumstix/gumstix-oe/com.gumstix.collection/packages/images/gumstix-minimal-image.bb, do_rootfs)
NOTE: package gumstix-minimal-image-1.0: started
NOTE: package gumstix-minimal-image-1.0-r0: task do_rootfs: started
NOTE: package gumstix-minimal-image-1.0-r0: task do_rootfs: completed
NOTE: package gumstix-minimal-image-1.0: completed
NOTE: Tasks Summary: Attempted 1299 tasks of which 1297 didn't need to
be rerun and 0 failed.
NOTE: build 200806102336: completed
bigger question is - would it not be a good option to steer away from
host tools and used native staged ones (standardize), especially for
packaging. If so then the (host) tar and ar commands are pretty widely
used currently.
I renamed my host /usr/bin/ar to /usr/bin/ar-old and built from scratch
watching what failed. most xxx-native packages seem to use (the host) ar
via libtool-native.
It would seem useful to add the binutils-native dependency to this
(libtool-native) recipe and use the natively staged ar for mostly
everything - especially given there is a known issue with at least the
Fedora 9 one when packaging.
And I guess this should go for tar as well where needed - other than
un-packaging source of course.
And how does this all fit in with the packaging_xxx classes and opkg?
Caveat: most of my testing of this is done on the somewhat older
gumstix-oe fork (which I have updated with the latest ipkg-utils,
gmp-4.2.2 and dbus-1.2.1 etc as required). I do also have a current dev
that I use - mostly for checking the latest recipes.
This seems however to be a current OE issue for stable, dev? and
gumstix-oe for Fedora 9 bleeding edgers.
Thanks and regards
Terry
BTW if I decide to take on a bug such as the above - should I change the
status to confirmed and assign to myself - not being a 'dev' and all?
^ permalink raw reply [flat|nested] 3+ messages in thread
* Re: tar & ar native vs host (Fedora 9 issues)
2008-06-10 11:50 tar & ar native vs host (Fedora 9 issues) Terry Kemp
@ 2008-06-11 1:07 ` Terry Kemp
2008-06-15 2:55 ` [RFC] " Mike (mwester)
0 siblings, 1 reply; 3+ messages in thread
From: Terry Kemp @ 2008-06-11 1:07 UTC (permalink / raw)
To: openembedded-devel
On Tue, 2008-06-10 at 23:50 +1200, Terry Kemp wrote:
> Hello devs,
>
> I have been working on
> http://bugs.openembedded.net/show_bug.cgi?id=4350
> and I have attached my recipe mods to this for review should someone
> care to look at them.
>
Just to confirm...
same deal with building latest mtn pull of dev...
[tmk@sulphur org.openembedded.dev]$ mtn status
Current branch: org.openembedded.dev
Changes against parent 30fc86774a3da064f530c39eeff1e97d6a7020a0
patched packages/linux/gumstix-kernel_2.6.21.bb
patched packages/linux/gumstix-linux.inc
patched site/common-glibc
[tmk@sulphur build]$
cat /build/OE/gumstix-connex/build/tmp/angstrom/work/gumstix-connex-angstrom-linux-gnueabi/minimal-image-1.0-r0/temp/log.do_rootfs.13586
+ rm -rf /build/OE/gumstix-connex/build/tmp/angstrom/rootfs
+ mkdir -p /build/OE/gumstix-connex/build/tmp/angstrom/rootfs
+ '[' 0 '!=' 1 ']'
+ for devtable
in /build/OE/org.openembedded.dev/files/device_table-minimal.txt
+ makedevs -r /build/OE/gumstix-connex/build/tmp/angstrom/rootfs
-D /build/OE/org.openembedded.dev/files/device_table-minimal.txt
+ rootfs_ipk_do_rootfs
+ set -x
+ package_update_index_ipk
+ set -x
+ ipkgarchs='all any noarch arm armv4 armv4t armv5e armv5te
gumstix-connex'
+ '[' '!' -z '' ']'
+
touch /build/OE/gumstix-connex/build/tmp/angstrom/deploy/glibc/ipk/Packages
+ ipkg-make-index
-r /build/OE/gumstix-connex/build/tmp/angstrom/deploy/glibc/ipk/Packages
-p /build/OE/gumstix-connex/build/tmp/angstrom/deploy/glibc/ipk/Packages
-l /build/OE/gumstix-connex/build/tmp/angstrom/deploy/glibc/ipk/Packages.filelist -m /build/OE/gumstix-connex/build/tmp/angstrom/deploy/glibc/ipk
+ for arch in '$ipkgarchs'
+ '['
-e /build/OE/gumstix-connex/build/tmp/angstrom/deploy/glibc/ipk/all/ ']'
+
touch /build/OE/gumstix-connex/build/tmp/angstrom/deploy/glibc/ipk/all/Packages
+ ipkg-make-index
-r /build/OE/gumstix-connex/build/tmp/angstrom/deploy/glibc/ipk/all/Packages -p /build/OE/gumstix-connex/build/tmp/angstrom/deploy/glibc/ipk/all/Packages -l /build/OE/gumstix-connex/build/tmp/angstrom/deploy/glibc/ipk/all/Packages.filelist -m /build/OE/gumstix-connex/build/tmp/angstrom/deploy/glibc/ipk/all/
Traceback (most recent call last):
File
"/build/OE/gumstix-connex/build/tmp/angstrom/staging/i686-linux/usr/bin/ipkg-make-index", line 114, in <module>
pkg = ipkg.Package(filename)
File
"/build/OE/gumstix-connex/build/tmp/angstrom/staging/i686-linux/usr/bin/ipkg.py", line 161, in __init__
tarStream = ar.open("control.tar.gz")
File
"/build/OE/gumstix-connex/build/tmp/angstrom/staging/i686-linux/usr/bin/arfile.py", line 59, in open
f = self._scan(fname)
File
"/build/OE/gumstix-connex/build/tmp/angstrom/staging/i686-linux/usr/bin/arfile.py", line 79, in _scan
size = int(descriptor[5])
IndexError: list index out of range
[tmk@sulphur build]$
Same fix being tested now.
Terry
^ permalink raw reply [flat|nested] 3+ messages in thread* [RFC] Re: tar & ar native vs host (Fedora 9 issues)
2008-06-11 1:07 ` Terry Kemp
@ 2008-06-15 2:55 ` Mike (mwester)
0 siblings, 0 replies; 3+ messages in thread
From: Mike (mwester) @ 2008-06-15 2:55 UTC (permalink / raw)
To: openembedded-devel
Barring objections, I'll submit the following patch to fix the bug
described below. Further details on the patch are in the bug report.
Regards,
Mike (mwester)
--- ipkg-utils/arfile.py.orig 2008-06-11 21:07:01.000000000 -0500
+++ ipkg-utils/arfile.py 2008-06-14 19:56:04.000000000 -0500
@@ -86,6 +86,8 @@
return FileSection(self.f, self.f.tell(), size)
# Skip data and loop
+ if size % 2:
+ size = size + 1
data = self.f.seek(size, 1)
# print hex(f.tell())
Terry Kemp wrote:
> On Tue, 2008-06-10 at 23:50 +1200, Terry Kemp wrote:
>> Hello devs,
>>
>> I have been working on
>> http://bugs.openembedded.net/show_bug.cgi?id=4350
>> and I have attached my recipe mods to this for review should someone
>> care to look at them.
>>
>
> Just to confirm...
>
> same deal with building latest mtn pull of dev...
>
> [tmk@sulphur org.openembedded.dev]$ mtn status
> Current branch: org.openembedded.dev
> Changes against parent 30fc86774a3da064f530c39eeff1e97d6a7020a0
> patched packages/linux/gumstix-kernel_2.6.21.bb
> patched packages/linux/gumstix-linux.inc
> patched site/common-glibc
>
>
> [tmk@sulphur build]$
> cat /build/OE/gumstix-connex/build/tmp/angstrom/work/gumstix-connex-angstrom-linux-gnueabi/minimal-image-1.0-r0/temp/log.do_rootfs.13586
> + rm -rf /build/OE/gumstix-connex/build/tmp/angstrom/rootfs
> + mkdir -p /build/OE/gumstix-connex/build/tmp/angstrom/rootfs
> + '[' 0 '!=' 1 ']'
> + for devtable
> in /build/OE/org.openembedded.dev/files/device_table-minimal.txt
> + makedevs -r /build/OE/gumstix-connex/build/tmp/angstrom/rootfs
> -D /build/OE/org.openembedded.dev/files/device_table-minimal.txt
> + rootfs_ipk_do_rootfs
> + set -x
> + package_update_index_ipk
> + set -x
> + ipkgarchs='all any noarch arm armv4 armv4t armv5e armv5te
> gumstix-connex'
> + '[' '!' -z '' ']'
> +
> touch /build/OE/gumstix-connex/build/tmp/angstrom/deploy/glibc/ipk/Packages
> + ipkg-make-index
> -r /build/OE/gumstix-connex/build/tmp/angstrom/deploy/glibc/ipk/Packages
> -p /build/OE/gumstix-connex/build/tmp/angstrom/deploy/glibc/ipk/Packages
> -l /build/OE/gumstix-connex/build/tmp/angstrom/deploy/glibc/ipk/Packages.filelist -m /build/OE/gumstix-connex/build/tmp/angstrom/deploy/glibc/ipk
>
> + for arch in '$ipkgarchs'
> + '['
> -e /build/OE/gumstix-connex/build/tmp/angstrom/deploy/glibc/ipk/all/ ']'
> +
> touch /build/OE/gumstix-connex/build/tmp/angstrom/deploy/glibc/ipk/all/Packages
> + ipkg-make-index
> -r /build/OE/gumstix-connex/build/tmp/angstrom/deploy/glibc/ipk/all/Packages -p /build/OE/gumstix-connex/build/tmp/angstrom/deploy/glibc/ipk/all/Packages -l /build/OE/gumstix-connex/build/tmp/angstrom/deploy/glibc/ipk/all/Packages.filelist -m /build/OE/gumstix-connex/build/tmp/angstrom/deploy/glibc/ipk/all/
> Traceback (most recent call last):
> File
> "/build/OE/gumstix-connex/build/tmp/angstrom/staging/i686-linux/usr/bin/ipkg-make-index", line 114, in <module>
> pkg = ipkg.Package(filename)
> File
> "/build/OE/gumstix-connex/build/tmp/angstrom/staging/i686-linux/usr/bin/ipkg.py", line 161, in __init__
> tarStream = ar.open("control.tar.gz")
> File
> "/build/OE/gumstix-connex/build/tmp/angstrom/staging/i686-linux/usr/bin/arfile.py", line 59, in open
> f = self._scan(fname)
> File
> "/build/OE/gumstix-connex/build/tmp/angstrom/staging/i686-linux/usr/bin/arfile.py", line 79, in _scan
> size = int(descriptor[5])
> IndexError: list index out of range
> [tmk@sulphur build]$
>
>
> Same fix being tested now.
>
>
> Terry
>
>
>
> _______________________________________________
> Openembedded-devel mailing list
> Openembedded-devel@lists.openembedded.org
> http://lists.linuxtogo.org/cgi-bin/mailman/listinfo/openembedded-devel
^ permalink raw reply [flat|nested] 3+ messages in thread
end of thread, other threads:[~2008-06-15 2:56 UTC | newest]
Thread overview: 3+ messages (download: mbox.gz follow: Atom feed
-- links below jump to the message on this page --
2008-06-10 11:50 tar & ar native vs host (Fedora 9 issues) Terry Kemp
2008-06-11 1:07 ` Terry Kemp
2008-06-15 2:55 ` [RFC] " Mike (mwester)
This is an external index of several public inboxes,
see mirroring instructions on how to clone and mirror
all data and code used by this external index.