multilib.bbclass: fix faulty redefinition of STAGING_KERNEL_DIR

Due to the problem fixed in
'56c677a multilib: Move redefinition of STAGING_DIR_KERNEL'
STAGING_KERNEL_DIR must be redefined for lib32 in multilib.bbclass.
However this redefinition expanded STAGING_KERNEL_DIR to an absolute
path. This unconsciously added the TMPDIR path in the sstate object,
causing packages depended on STAGING_KERNEL_DIR being rebuild if the
TMPDIR was changed.

Solve this by forcing the unexpanded TMPDIR variable to remain in the
beginning of STAGING_DIR_KERNEL (as default). Since TMPDIR is included in
BB_HASHBASE_WHITELIST, the sstate object will not be depended on the
expanded path anymore.

Signed-off-by: Petter Mabäcker <petter@technux.se>
Signed-off-by: Ross Burton <ross.burton@intel.com>

diff --git a/meta/classes/multilib.bbclass b/meta/classes/multilib.bbclass
index ab04597f93e3c8dee5b013928afda7dd5ad971d9..816f54e7f5d10bce1adbb4407dd77c3ff63db36a 100644 (file)
--- a/meta/classes/multilib.bbclass
+++ b/meta/classes/multilib.bbclass
@@ -4,7 +4,9 @@ python multilib_virtclass_handler () {
     if cls != "multilib" or not variant:
         return
 
-    e.data.setVar('STAGING_KERNEL_DIR', e.data.getVar('STAGING_KERNEL_DIR'))
+    localdata = bb.data.createCopy(e.data)
+    localdata.delVar('TMPDIR')
+    e.data.setVar('STAGING_KERNEL_DIR', localdata.getVar('STAGING_KERNEL_DIR'))
 
     # There should only be one kernel in multilib configs
     # We also skip multilib setup for module packages.