Thoughts on current OE gta01/gta02 integration, kernel, uboot
Harald Welte
laforge at openmoko.org
Wed Oct 3 21:32:24 CEST 2007
Hi!
Today I've finally started to use 2007.2 and I have the following
remarks:
1) the yaffs2 support is not working since there's a missing parameter
to the mkyaffs2image in bitbake.conf. I'm not yet sure which one is
the right one for us, will submit a patch later on.
2) the multi-mach support semems to work fine in general, however:
a) when starting from MACHINE="fic-gta01" and then changeing to
gta02, it rebuilds quite a number of 'generic' packages such as
gtk+. I don't know why that is the case.
b) a couple of task-*_fic-gta02.ipk files end up in
deploy/glibc/ipk/all. I don't think that's right.
c) the u-boot recipe seems completely broken, since building
uboot-openmoko for fic-gta02 ends up building all the gta01
variants of u-boot, but placed in the fic-gta02 directory, which
is not good.
d) alsa-states_0.1.0-r3_fic-gta02.ipk and
sysvinit-inittab_2.86-r36_fic-gta02.ipk are put in
deploy/glibc/ipk/armv4t. Is that correct?
e) the entire kernel is re-built, which is not neccessarry. We have
(and will continue to have) one kernel binary that runs on a
variety of machines. Our current kernel should run on qt2410,
smdk2410, smdk2440, fic-gta01 and fic-gta02. I don't know what
bb/OE magic is required to teach the buildsystem about this.
f) dfu-util is built for fic-gta01 but not for fic-gta02. It should
be built for both, but actually not placed into the images
directory. It's a native binary, and it's completely independent
of the target machine arch.
g) sjf2410 is built for fic-gta02 but not for fic-gta01. I think we
don't need it built by default on either of them.
h) I think we need a meta-task that builds the entire system, i.e.
"openmoko-image + uboot-openmoko" or
"openmoko-devel-image + uboot-openmoko". I think there should be
a single command that enables the user to rebuild all software.
i) the modules-*.tgz file doesn't contain the patchset svn revision
number
j) the following packages were built in fic-gta02 for yet unknown
reason. I personally don't believe there's something gta02
specific in them:
pointercal-dbg_0.0-r4_fic-gta02.ipk
pointercal_0.0-r4_fic-gta02.ipk
pointercal-dev_0.0-r4_fic-gta02.ipk
usb-gadget-mode-dbg_0.0.2-r0_fic-gta02.ipk
usb-gadget-mode_0.0.2-r0_fic-gta02.ipk
usb-gadget-mode-dev_0.0.2-r0_fic-gta02.ipk
keymaps-dbg_1.0-r14_fic-gta02.ipk
keymaps_1.0-r14_fic-gta02.ipk
keymaps-dev_1.0-r14_fic-gta02.ipk
ipkg-collateral-dbg_1.0-r7_fic-gta02.ipk
ipkg-collateral_1.0-r7_fic-gta02.ipk
ipkg-collateral-dev_1.0-r7_fic-gta02.ipk
netbase-dbg_4.21-r23_fic-gta02.ipk
netbase_4.21-r23_fic-gta02.ipk
netbase-doc_4.21-r23_fic-gta02.ipk
netbase-dev_4.21-r23_fic-gta02.ipk
gtk+_2.10.14-r7_fic-gta02.ipk (and tons of other gtk+ packages)
gdk-pixbuf-loader-png_2.10.14-r7_fic-gta02.ipk
gdk-pixbuf-loader-jpeg_2.10.14-r7_fic-gta02.ipk
gdk-pixbuf-loader-ico_2.10.14-r7_fic-gta02.ipk
gdk-pixbuf-loader-gif_2.10.14-r7_fic-gta02.ipk
gdk-pixbuf-loader-xpm_2.10.14-r7_fic-gta02.ipk
gdk-pixbuf-loader-wbmp_2.10.14-r7_fic-gta02.ipk
gdk-pixbuf-loader-ras_2.10.14-r7_fic-gta02.ipk
gdk-pixbuf-loader-pnm_2.10.14-r7_fic-gta02.ipk
gdk-pixbuf-loader-bmp_2.10.14-r7_fic-gta02.ipk
gdk-pixbuf-loader-ani_2.10.14-r7_fic-gta02.ipk
gtk-immodule-xim_2.10.14-r7_fic-gta02.ipk
gtk-immodule-am-et_2.10.14-r7_fic-gta02.ipk
gdk-pixbuf-loader-xbm_2.10.14-r7_fic-gta02.ipk
gdk-pixbuf-loader-tga_2.10.14-r7_fic-gta02.ipk
gdk-pixbuf-loader-pcx_2.10.14-r7_fic-gta02.ipk
gtk-immodule-ti-er_2.10.14-r7_fic-gta02.ipk
gtk-immodule-thai_2.10.14-r7_fic-gta02.ipk
gtk-immodule-ipa_2.10.14-r7_fic-gta02.ipk
gtk-immodule-inuktitut_2.10.14-r7_fic-gta02.ipk
gtk-immodule-cyrillic-translit_2.10.14-r7_fic-gta02.ipk
gtk-immodule-cedilla_2.10.14-r7_fic-gta02.ipk
gtk-printbackend-lpr_2.10.14-r7_fic-gta02.ipk
gtk-printbackend-file_2.10.14-r7_fic-gta02.ipk
gtk-printbackend-cups_2.10.14-r7_fic-gta02.ipk
gtk-immodule-viqr_2.10.14-r7_fic-gta02.ipk
gtk-immodule-ti-et_2.10.14-r7_fic-gta02.ipk
neod_0.1.0+svnr2956-r1_fic-gta02.ipk
neod-dev_0.1.0+svnr2956-r1_fic-gta02.ipk
neod-dbg_0.1.0+svnr2956-r1_fic-gta02.ipk
It would be great if Graeme or somebody else with bitbake+OE background
could look into resolving those issues.
3) svn revisions for kernel + u-boot are frozen right now. I don't say
thre is no reason to do so, but I would want to know why those
specific revisions have been selected, and what particular
regressions are worked around by using old versions.
Cheers,
--
- Harald Welte <laforge at openmoko.org> http://openmoko.org/
============================================================================
Software for the world's first truly open Free Software mobile phone
More information about the distro-devel
mailing list