Carsten Schlote wrote: > Moin Marc > >>> Carsten Schlote (15): >>> [ipkg-push] Create stage file for ipkg-push make target >>> [images] Creation of 'Packages', removed redirection to nil: >>> [host-localedef] Added missing get make target >>> [dtc] Added option to install oftree into /boot >>> [kernel] Fixed creation of boot directory using appropriate >>> rights >> I think I've cherry-picked these ^^ > > So I won't touch them again. I tested the DTC and kernel stuff and > it works fine - I can update the whole system with ipkg update and on > next reboot the latest kernel and oftree are loaded from /boot.
cool :)
you can just rebase your tree to lastest master.
> BTW: Any plans to update the phytec-mpc5200b-{tiny|...} bootloader (also
> and preferably u-boot-v2)? Or any started work on it, which I could pick
it's called barebox now
> up and continue? I'd like to fetch the kernel and oftree from either
> ext2/3/4 or jffs2 and get rid of the extra partitions on CF and the two
> extra MTD parts....
We have no filesystem supprt, but you're welcome to contribute!
>>> [images] Added option IMAGE_INSTALL_FROM_IPKG_REPOSITORY
>>> [images] Fixup ipkg status file in worktmp for image creation
>
> I'm testing the two modes of operation right now and check that
> ipkg/status is matching the packages, which were used to create the
> images.
We're planing to make a release quite soon, this will go into my next
branch as soon as I have time and I'm not as tiered as I was while
cherry-picking the other (more trivial) patches :)
>>> [images] Reworked creation IPKG_FILES and
> $(IMGDIR)/permissions
>
> Here is one minor open issue: I currently collect the ipk files without
> caring about the version. As long as no old ipk files are around and no
> upstream update causes a new package version, this is fine. But with
> multiple files (e.g. foo_1.0{0|1|...}_arch.ipk) the wrong file for image
> creation might be collected (first hit is taken).
Hmmm....what about deleting "all" foo_*.ipk directly before generating a
new one? E.g.: in the install finish routine?
> So I fear I have to add a ptxd library function to consistently convert
> a PKG_VERSION into a IPKG_VERSION and use it at several place to change
> the match patterns to include it as well.
>
> For the IPKG_VERSION we need in addition to the '-_'->'.' conversion
> some code to
> replace letters by numbers, which can be parsed and compared correctly
> and result into the right answer for the "Which package is newer?"
> question.
sounds like a plan.
>>> [kernel] Use real imagename instead of linuximage, add legacy
> link
>> I have to test if the link works in all cases. Think of someone
> copying
>> the kernel image with "cp -a"
>
> Right, there might be some issues. But it was intended as some kind of
> legacy support - at least until we changed all our boxes to the 'new'
> name. I really prefer to be able to distinguish between linux images by
> their names. Instead of 'linuximage' you setup the correct name in
> uboot/PXE/dhcp - no extra work, less ambigious.
what about two plain files, the "leagacy" linuximage and the actual
image name?
> The legacy link stuff definitivly breaks, if your /boot partition is FAT
> and symbolic links can't be used. For TFTP/NFS setup the link should
> work as long as ptxdist is used to update and create the kernel image
> and the legacy link.
What you patch does is:
-install -m 644 $$i $(IMAGEDIR)/linuximage;
+install -m 644 $$i $(IMAGEDIR)/$(call remove_quotes,$(KERNEL_IMAGE));
[...]
+ln -sf $(KERNEL_IMAGE) $(IMAGEDIR)/linuximage
It changes the kernel images's name and creates a link in IMAGEDIR. I
don't see the connection to "/boot" on the target.
>
>>> [vpnc][v4] Added vpnc 0.5.3 package, a VPN client for Cisco
>>> concentrators
>> vpnc still does not work here
>
> We compiled it on several Ubuntu machines last night. Some are matured,
> others newly setup for development - no problems. So I'm a bit glueless,
> why it doesn't compile on your box.
have you read Wolfram's comment, that the patch was missing some lines...
> I attached the logfile for a vpnc package rebuild - I really can't see
> any problems, the makeman.pl script correctly handles the
> cross-compilation issue.
>
> BTW: We have the host-perl (5.10.1) package enabled. It is required for
> the target-perl build process. The perl binaries in sysroot-host overlay
> our host machines' perl binaries - maybe this is the reason why it works
> here?
hmmmm....at least that's suspicious
cheers, Marc
--
Pengutronix e.K. | Marc Kleine-Budde |
Industrial Linux Solutions | Phone: +49-231-2826-924 |
Vertretung West/Dortmund | Fax: +49-5121-206917-5555 |
Amtsgericht Hildesheim, HRA 2686 | http://www.pengutronix.de |
signature.asc
Description: OpenPGP digital signature
-- ptxdist mailing list [email protected]
