[OE-core] [PATCH] pango: have postinstalls run at do_rootfs time

Laurentiu Palcu laurentiu.palcu at intel.com
Thu Dec 20 16:21:27 UTC 2012



On 12/20/2012 05:59 PM, Burton, Ross wrote:
> On 20 December 2012 15:21, Phil Blundell <philb at gnu.org> wrote:
>> On Wed, 2012-12-19 at 18:03 +0200, Laurentiu Palcu wrote:
>>> Since pango-native is built anyway and all the modules are in the native
>>> sysroot, create the cache file by scanning those files instead of the
>>> target files. The latter will fail because the shared objects wouldn't
>>> be from the same ELF class.
>>
>> Is it guaranteed that the native build will have the same set of
>> modules?  It seems to me that it would be safer to scan the target ones
>> under qemu as you did for gtk-immodules.
> 
> I suspected there was no such real thing as an out-of-tree Pango
> module, but I was wrong (http://graphite.sil.org/).
But, in this case, we could use the qemu method only for the graphite
package postinstall... We don't have to use it for pango too.

> 
> Digging quickly into the Debian packaging, they generate the modules
> list at build time (not in a postinst) by running pango-querymodules
> over the directories they just built, filtering the paths (this
> happens in dh_pangomodules).  They also generate a modules list file
> per package and ship it, and pango reads the multiple files.  I'm not
> sure if this is upstream behaviour or a Debian-specific patch.
How do they handle the cross-compiling issue? They cannot run the native
pango-querymodules on a cross-compiled shared object. It will fail.

Thanks,
Laurentiu
> 
> Ross
> 




More information about the Openembedded-core mailing list