Re: useradd changes to dbus, base-passwd

["Eric Bénard" <eric@eukrea.com>]

Hi Scott,

Le 07/11/2011 00:16, Scott Garman a écrit :
> This looks good, thank you! Two questions though:
>
> When I was testing my own variation of this fix out, I found that testing
> bb.data.inherits_class('nativesdk', d) did not work as I expected. It returned
> false for a recipe I was testing with that used BBCLASSEXTEND = "native
> nativesdk". So instead I was checking if PN ended with -nativesdk. Are you
> certain the inherits_class test works?
>
this seems to work as expected in the present case :

$ grep BBCLASSEXTEND sources/openembedded-core/meta/recipes-core/dbus/dbus.inc
BBCLASSEXTEND = "native nativesdk"

$ bitbake dbus -e |grep ^DEPENDS=
DEPENDS="pkgconfig-native autoconf-native automake-native libtool-native 
libtool-cross gnu-config-native virtual/gettext gettext-native 
virtual/arm-angstrom-linux-gnueabi-gcc 
virtual/arm-angstrom-linux-gnueabi-compilerlibs virtual/libc  expat 
virtual/libintl virtual/libx11 libsm shadow-native shadow-sysroot base-passwd 
update-rc.d-native"

$ bitbake dbus-native -e |grep ^DEPENDS=
DEPENDS="pkgconfig-native autoconf-native automake-native libtool-native 
gnu-config-native gettext-minimal-native  expat-native virtual/libintl-native 
shadow-native shadow-sysroot base-passwd update-rc.d-native"

$ bitbake dbus-nativesdk -e |grep ^DEPENDS=
DEPENDS="pkgconfig-native autoconf-native automake-native libtool-native 
libtool-cross gnu-config-native virtual/gettext-nativesdk gettext-native 
virtual/i686-angstromsdk-linux-gcc-crosssdk 
virtual/i686-angstromsdk-linux-compilerlibs-nativesdk virtual/libc-nativesdk 
expat-nativesdk virtual/libintl-nativesdk virtual/libx11-nativesdk 
update-rc.d-native"

$ bitbake dbus-nativesdk -e |grep ^SSTATEPOSTIN
$ bitbake dbus-native -e |grep ^SSTATEPOSTIN
SSTATEPOSTINSTFUNCS="useradd_sysroot_sstate"
$ bitbake dbus -e |grep ^SSTATEPOSTIN
SSTATEPOSTINSTFUNCS="useradd_sysroot_sstate"


> My second question is why the autotools_do_install is needed for the nativesdk
> case?
>
because without the do_install_virtclass-nativesdk() , that's the standard 
do_install which is used which tries to setup some rigths using chown on the 
files and that fails (and is not very useful for a nativesdk).
Maybe we could use a similar do_install as the one used for native.

Eric