Koen Kooi [Sat, 5 May 2007 16:34:06 +0000 (16:34 +0000)]
insane.bbclass: on check for references to WORKDIR in RPATH, instead of WORKDIR and STAGING
[18:23] likewise: would be my proposed fix for now, until we figure out if cross tooling need to RPATH to it's own libs.
classes/magicbox-image.bbclass: Add a class that handles the generation of flash images for Magicbox
It creates "ready-to-burn" squashfs and squashfs-lza images and a single file kernel+jffs2 image
Jamie Lenehan [Fri, 4 May 2007 02:09:42 +0000 (02:09 +0000)]
perl: Add the .debug directories that to FILES_${PN}-dbg in the
cpan-base.bbclass, and then update the cpan modules to no longer manually
specify this. Also remove references to the new removed .packlist files.
Jamie Lenehan [Fri, 4 May 2007 01:47:32 +0000 (01:47 +0000)]
perl-native 5.8.8: Apply the patch from perl to stop the .packlist files
from being installed. Since we use perl-native to do the cpan module
installs we need to changed perl-native as well to stop cpan modules from
getting .packlist files. This is a new patch since just the .packlist
related part has been extracted from the larger patch.
Jamie Lenehan [Fri, 4 May 2007 01:34:31 +0000 (01:34 +0000)]
perl/perl-native 5.8.8: Move the creation of the hostperl link in staging
from the configure task in perl to the stage task in perl-native where it
belongs - perl shouldn't be messing with the staged perl-native.
Jamie Lenehan [Fri, 4 May 2007 01:24:59 +0000 (01:24 +0000)]
perl: Extract common functionality from cpan and cpan_build classes into a
cpan-base class and update cpan_build to work with the new perl layout that
was added with perl 5.8.8.
Paul Sokolovsky [Thu, 3 May 2007 12:09:21 +0000 (12:09 +0000)]
libqpe-opie cvs: Unbreak logging.
* Again, it's bad idea to randomly disable some logging sources based on
obscure compile-time defines (QT_DEBUG is not mentioned in qte-2.3.10, so
how could it get such name at all?). OE OPIE is going to have logging facility
helping to resolve issues, not obfuscate them.
Paul Sokolovsky [Thu, 3 May 2007 05:21:22 +0000 (05:21 +0000)]
opie-taskbar cvs: Don't bother to start qss from C++ code.
* Because it's useless - to have a separate process for sound server,
and then have mishacks to start/keep alive it hardcoded in C++. That doesn't
work somehow (h4000, hx4700, etc. machines have issues), and then it
completely non-debuggable.
* So instead, qss should be started separately, of course from a shell script.
* Partially fixes #2211.
packages/gcc/gcc_4.1.1.bb: Do not apply ppc soft-float patches when building with uclibc
This is a dirty hack to get gcc 4.1.1 to compile for glibc AND uclibc on ppc
the patch that is need it to get gcc support soft-floats with glibc, makes gcc fail with uclibc
Jamie Lenehan [Wed, 2 May 2007 05:58:32 +0000 (05:58 +0000)]
net-snmp 5.1.2: Fix for do_configure hanging when using bitbake
1.8.2. We need to pass "--with-defaults" to configure tell it we want
the defaults so it doesn't warn that the defaults are going to be
used and ask them to confirm each of the defaults.
Jamie Lenehan [Wed, 2 May 2007 02:14:14 +0000 (02:14 +0000)]
cpan.bbclass: Stop the cpan modules which ask you to confirm the
configuration from hanging bitbake. They expect you to press y to indicate
that you are ok with the configuration. For some reason there was no problem
in bitbake 1.6 but this caused hangs in do_configure with bitbake 1.8.
Jamie Lenehan [Wed, 2 May 2007 02:10:55 +0000 (02:10 +0000)]
cpan.bbclass: Building cpan modules needs to depend on both perl and
perl-native. They are built with perl-native but use the
configuration information from perl. This issue wasn't showing up
with bitbake 1.6 but is with 1.8.2 where perl is configured and
compiled perl but not staged prior to moving on.
Koen Kooi [Tue, 1 May 2007 16:54:35 +0000 (16:54 +0000)]
glibc 2.5: fix the strlen() issues on big endian EABI, thanks to Lennert for tracking down the culprit
* This is not a proper fix, we still need to find out wht __ARMEB__ isn't defined
Jamie Lenehan [Tue, 1 May 2007 16:21:30 +0000 (16:21 +0000)]
perl 5.8.8: Fix to tell bitbake about the perl module packages. I'd left out
the PACKAGES_DYNAMIC declaration for the perl modules. This declaration
appears to be global though - so if either of the older perl versions exist
it actually works - bitbake picks up the definition from the older recipes
even though it isn't building them. This shows up as "No providers of
runtime build target <x>" errors where x is a perl module. Thanks to Koen
for figuring this one out.
Koen Kooi [Tue, 1 May 2007 12:20:18 +0000 (12:20 +0000)]
xtscal: reinstate tr toupper filter so everything works again.
NOTE: it would be nice if machine maintainers run the script on their device instead of pasting a line from /proc/cpuinfo. If the htc universal people had done so, they would have noticed everything comes up in *uppercase*. So xtscal wasn't broken
Jamie Lenehan [Tue, 1 May 2007 11:51:40 +0000 (11:51 +0000)]
perl 5.8.8: Allow perl to build when using an external toolchain.
This is done by allowing gcc to search for errno.h by itself instead
of manually searching for. The manual search was looking in
STAGING_INCDIR and that's not where the external toolchains headers
are. This whole test is really for handle other compilers and
operating systems, so the simple make gcc do itself should be fine
for us. Closes #1700.