[PATCH 1/2] oeqa/runtime/multilib: add test for libc

[PATCH 1/2] oeqa/runtime/multilib: add test for libc

[Ross Burton]

Add a basic test to verify that /lib/libc.so.6 and /lib32/libc.so.6 have the
right ELF class.

Signed-off-by: Ross Burton <ross.burton@intel.com>
---
 meta/lib/oeqa/runtime/multilib.py | 19 ++++++++++++++++++-
 1 file changed, 18 insertions(+), 1 deletion(-)

diff --git a/meta/lib/oeqa/runtime/multilib.py b/meta/lib/oeqa/runtime/multilib.py
index 1c1729b..e1bcc42 100644
--- a/meta/lib/oeqa/runtime/multilib.py
+++ b/meta/lib/oeqa/runtime/multilib.py
@@ -20,8 +20,25 @@ class MultilibTest(oeRuntimeTest):
         else:
             self.fail("Cannot parse readelf output\n" + s)
 
-    @testcase('279')
     @skipUnlessPassed('test_ssh')
+    def test_check_multilib_libc(self):
+        """
+        Check that a multilib image has both 32-bit and 64-bit libc in.
+        """
+
+        (status, output) = self.target.run("readelf -h /lib/libc.so.6")
+        self.assertEqual(status, 0, "Failed to readelf /lib/libc.so.6")
+        class32 = self.parse(output)
+
+        (status, output) = self.target.run("readelf -h /lib64/libc.so.6")
+        self.assertEqual(status, 0, "Failed to readelf /lib64/libc.so.6")
+        class64 = self.parse(output)
+
+        self.assertEqual(class32, "ELF32", msg="/lib/libc.so.6 isn't ELF32 (is %s)" % class32)
+        self.assertEqual(class64, "ELF64", msg="/lib64/libc.so.6 isn't ELF64 (is %s)" % class64)
+
+    @testcase('279')
+    @skipUnlessPassed('test_check_multilib_libc')
     def test_file_connman(self):
         self.assertTrue(oeRuntimeTest.hasPackage('lib32-connman-gnome'), msg="This test assumes lib32-connman-gnome is installed")
 
-- 
2.1.4

[PATCH 2/2] oeqa/runtime/multilib: change test case to use glib-2.0-utils

[Ross Burton]

Using connman-applet as a multilib test case seems quite heavy due to the
numerous dependencies it has, whereas the same test ("do binaries get swapped
correctly") can be done with a lower-level library containing binaries, for
example glib-2.0.

Signed-off-by: Ross Burton <ross.burton@intel.com>
---
 meta/lib/oeqa/runtime/multilib.py | 10 +++++-----
 1 file changed, 5 insertions(+), 5 deletions(-)

diff --git a/meta/lib/oeqa/runtime/multilib.py b/meta/lib/oeqa/runtime/multilib.py
index e1bcc42..59ae86d 100644
--- a/meta/lib/oeqa/runtime/multilib.py
+++ b/meta/lib/oeqa/runtime/multilib.py
@@ -39,10 +39,10 @@ class MultilibTest(oeRuntimeTest):
 
     @testcase('279')
     @skipUnlessPassed('test_check_multilib_libc')
-    def test_file_connman(self):
-        self.assertTrue(oeRuntimeTest.hasPackage('lib32-connman-gnome'), msg="This test assumes lib32-connman-gnome is installed")
+    def test_file_glib(self):
+        self.assertTrue(oeRuntimeTest.hasPackage('lib32-libglib-2.0-utils'), msg="This test assumes lib32-libglib-2.0-utils is installed")
 
-        (status, output) = self.target.run("readelf -h /usr/bin/connman-applet")
-        self.assertEqual(status, 0, "Failed to readelf /usr/bin/connman-applet")
+        (status, output) = self.target.run("readelf -h /usr/bin/glib-genmarshal")
+        self.assertEqual(status, 0, "Failed to readelf /usr/bin/glib-genmarshal")
         theclass = self.parse(output)
-        self.assertEqual(theclass, "ELF32", msg="connman-applet isn't ELF32 (is %s)" % theclass)
+        self.assertEqual(theclass, "ELF32", msg="glib-genmarshal isn't ELF32 (is %s)" % theclass)
-- 
2.1.4

Re: [PATCH 2/2] oeqa/runtime/multilib: change test case to use glib-2.0-utils

[Burton, Ross]

[-- Attachment #1: Type: text/plain, Size: 502 bytes --]

On 29 August 2015 at 00:43, Ross Burton <ross.burton@intel.com> wrote:

> Using connman-applet as a multilib test case seems quite heavy due to the
> numerous dependencies it has, whereas the same test ("do binaries get
> swapped
> correctly") can be done with a lower-level library containing binaries, for
> example glib-2.0.
>

Updating the related bug made me realise that this would mean adding logic
to the autobuilder to handle new and old tests, so it really isn't worth it.

Ross

[-- Attachment #2: Type: text/html, Size: 922 bytes --]

Re: [PATCH 2/2] oeqa/runtime/multilib: change test case to use glib-2.0-utils

[Paul Eggleton]

On Saturday 29 August 2015 00:54:10 Burton, Ross wrote:
> On 29 August 2015 at 00:43, Ross Burton <ross.burton@intel.com> wrote:
> > Using connman-applet as a multilib test case seems quite heavy due to the
> > numerous dependencies it has, whereas the same test ("do binaries get
> > swapped
> > correctly") can be done with a lower-level library containing binaries,
> > for
> > example glib-2.0.
> 
> Updating the related bug made me realise that this would mean adding logic
> to the autobuilder to handle new and old tests, so it really isn't worth it.

Why is that? Really the autobuilder should just be running all valid tests for 
the image and not have any knowledge of what tests are being run.

Cheers,
Paul

-- 

Paul Eggleton
Intel Open Source Technology Centre

Re: [PATCH 2/2] oeqa/runtime/multilib: change test case to use glib-2.0-utils

[Burton, Ross]

[-- Attachment #1: Type: text/plain, Size: 561 bytes --]

On 29 August 2015 at 12:35, Paul Eggleton <paul.eggleton@linux.intel.com>
wrote:

> Why is that? Really the autobuilder should just be running all valid tests
> for
> the image and not have any knowledge of what tests are being run.
>

Because the autobuilder needs to build specific images: the multilib runs
build a core-image-sato but add IMAGE_INSTALL_append = "
lib32-connman-gnome" to the auto.conf to get the right bits in.

Merging this would mean a version bump and the AB respecting the version as
to what libraries it installs.

Ross

[-- Attachment #2: Type: text/html, Size: 1044 bytes --]