[oe] [PATCH 2/2] glibc-package: set LD_LIBRARY_PATH on binary locale generation

Phil Blundell pb at reciva.com
Fri May 15 19:44:50 UTC 2009


On Fri, 2009-05-15 at 23:13 +0400, Roman I Khimov wrote:
> On Friday 15 May 2009 19:32:14 Phil Blundell wrote:
> > On Fri, 2009-05-15 at 17:27 +0400, Roman I Khimov wrote:
> > > BTW, while at glibc-package, is there any reason for
> > > BINARY_LOCALE_ARCHES? We're using binary locale generation even for
> > > x86_64 and it is useful for us there. There is
> > > ENABLE_BINARY_LOCALE_GENERATION switch already to control whether one
> > > wants binary locales generated and if one does, why not on all
> > > architectures?
> >
> > BINARY_LOCALE_ARCHES is intended to stop glibc-package from trying to
> > generate the locales for targets where this is known in advance to be a
> > losing proposition (generally because qemu doesn't support the target
> > processor family).  If we didn't do that, since
> > ENABLE_BINARY_LOCALE_GENERATION defaults to on, people trying to build
> > for non-qemu-enabled targets would just get some cryptic error during
> > the packaging process.
> 
> OK, then I guess we should extend it at least with "i?86" and "x86_64", 
> because that work perfect with QEMU.

Yes, that'd be a fine plan. 

In an ideal world, you'd arrange for each version of qemu-native to
PROVIDE "virtual/qemu-native-TARGET-SYS" for each architecture that the
version in question supported, then make glibc RECOMMEND
virtual/qemu-native-${TARGET_SYS} and enable binary locales by default
if that recommendation was satisfied at runtime.  But, obviously,
bitbake doesn't currently do RECOMMENDS so some infrastructure would be
needed to make this work.

p.






More information about the Openembedded-devel mailing list