From: Tim Gore <[email protected]>

Currently when IGT is built for Android the resulting test
executables go to /system/bin, which is not ideal. After
discussion with the core validation team i have moved them
to /system/vendor/intel/validation/core/igt by setting
LOCAL_MODULE_PATH.

I have also added a --defsym linker option to export a
symbol that allows a script to easily distinguish between
tests that have subtests and those that dont. There are
better ways to do this (viz, in the source code) but
because the igt tests are not written consistently this
would require many more changes.

Signed-off-by: Tim Gore <[email protected]>
---
 tests/Android.mk | 10 ++++++++--
 1 file changed, 8 insertions(+), 2 deletions(-)

diff --git a/tests/Android.mk b/tests/Android.mk
index db4a4aa..ad61ab8 100644
--- a/tests/Android.mk
+++ b/tests/Android.mk
@@ -18,6 +18,9 @@ define add_test
     LOCAL_SHARED_LIBRARIES := ${IGT_LOCAL_SHARED_LIBRARIES}
 
     LOCAL_MODULE_TAGS := optional
+    # ask linker to define a specific symbol; we use this to identify IGT tests
+    LOCAL_LDFLAGS := -Wl,--defsym=$2=0
+    LOCAL_MODULE_PATH := 
$(ANDROID_PRODUCT_OUT)/$(TARGET_COPY_OUT_VENDOR)/intel/validation/core/igt
 
     include $(BUILD_EXECUTABLE)
 endef
@@ -68,7 +71,10 @@ else
     IGT_LOCAL_CFLAGS += -DANDROID_HAS_CAIRO=0
 endif
 
-tests_list := $(filter-out $(skip_tests_list),$(TESTS_progs) $(TESTS_progs_M) 
$(HANG) $(TESTS_testsuite))
+# create two test lists, one for simple single tests, one for tests that have 
subtests
+tests_list   := $(filter-out $(skip_tests_list),$(TESTS_progs) $(HANG) 
$(TESTS_testsuite))
+tests_list_M := $(filter-out $(skip_tests_list),$(TESTS_progs_M))
 
-$(foreach item,$(tests_list),$(eval $(call add_test,$(item))))
+$(foreach item,$(tests_list),$(eval $(call 
add_test,$(item),"IGT_SINGLE_TEST")))
+$(foreach item,$(tests_list_M),$(eval $(call 
add_test,$(item),"IGT_MULTI_TEST")))
 
-- 
1.9.2

_______________________________________________
Intel-gfx mailing list
[email protected]
http://lists.freedesktop.org/mailman/listinfo/intel-gfx

Reply via email to