Skip to contentSkip to conversation

Patch review

PATCH In review Steve Sakoman 2 participants 7 messages Oct 10, 2024 – Oct 17, 2024

lore
Share & tools
lore
#0 lore raw

cover Patch review

Please review this set of changes for kirkstone and have comments back by end of day Friday, October 11

Passed a-full on autobuilder:

https://valkyrie.yoctoproject.org/#/builders/29/builds/246

The following changes since commit 4ef9b4d947b6e8ed7637cdd9e722c3f0c83990dd:

zstd: fix LICENSE statement (2024-10-08 11:49:12 -0700)

are available in the Git repository at:

https://git.openembedded.org/openembedded-core-contrib stable/kirkstone-nut
https://git.openembedded.org/openembedded-core-contrib/log/?h=stable/kirkstone-nut

Steve Sakoman (2):
valgrind: disable avx_estimate_insn.vgtest selftest/runtime_test/virgl: Disable for all fedora

 meta/lib/oeqa/selftest/cases/runtime_test.py      | 2 +-
 meta/recipes-devtools/valgrind/valgrind_3.18.1.bb | 8 ++++++++
 2 files changed, 9 insertions(+), 1 deletion(-)
Signature · 1 line
2.34.1
#1 lore raw

PATCH 1/2 valgrind: disable avx_estimate_insn.vgtest

This test fails consistently on the new valkyrie autobuilder cluster

The estimate instructions (rcpss, rcpps, rsqrtps, rsqrtss) are, as the name suggests, not expected to give a fully accurate result. They may produce slighly different results on different CPU families because their results are not defined by the IEEE standard.

---
 meta/recipes-devtools/valgrind/valgrind_3.18.1.bb | 8 ++++++++
 1 file changed, 8 insertions(+)
meta/recipes-devtools/valgrind/valgrind_3.18.1.bb +8 -0
index 0c297ad6b9..2a1bf19da5 100644
@@ -237,6 +237,14 @@ do_install_ptest() {
     # This is known failure see https://bugs.kde.org/show_bug.cgi?id=435732
     rm ${D}${PTEST_PATH}/memcheck/tests/leak_cpp_interior.vgtest
 
+    # This test fails consistently on the new valkyrie autobuilder cluster
+    # It tests the avx estimate instructions
+    # The estimate instructions (rcpss, rcpps, rsqrtps, rsqrtss) are, as the
+    # name suggests, not expected to give a fully accurate result. They may
+    # produce slighly different results on different CPU families because
+    # their results are not defined by the IEEE standard
+    rm ${D}${PTEST_PATH}/none/tests/amd64/avx_estimate_insn.vgtest
+
     # As the binary isn't stripped or debug-splitted, the source file isn't fetched
     # via dwarfsrcfiles either, so it needs to be installed manually.
     mkdir -p ${D}/usr/src/debug/${PN}/${EXTENDPE}${PV}-${PR}/${BP}/none/tests/
Signature · 1 line
2.34.1
#2 lore raw

PATCH 2/2 selftest/runtime_test/virgl: Disable for all fedora

In reply to Steve Sakoman

We can't support vgem on RHEL derived distros so disable this test for all fedora hosts rather than specific versions.

---
 meta/lib/oeqa/selftest/cases/runtime_test.py | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)
meta/lib/oeqa/selftest/cases/runtime_test.py +1 -1
index 7dcdfd0ab2..93c6da1889 100644
@@ -258,7 +258,7 @@ class TestImage(OESelftestTestCase):
 
         distro = oe.lsb.distro_identifier()
         if distro and (distro in ['debian-9', 'debian-10', 'centos-7', 'centos-8', 'ubuntu-16.04', 'ubuntu-18.04'] or
-            distro.startswith('almalinux') or distro.startswith('rocky')):
+            distro.startswith('almalinux') or distro.startswith('fedora') or distro.startswith('rocky')):
             self.skipTest('virgl headless cannot be tested with %s' %(distro))
 
         render_hint = """If /dev/dri/renderD* is absent due to lack of suitable GPU, 'modprobe vgem' will create one suitable for mesa llvmpipe software renderer."""
Signature · 1 line
2.34.1
#3 lore raw

Re: [OE-core][kirkstone 2/2] selftest/runtime_test/virgl: Disable for all fedora

Quoted 3 lines · Steve Sakoman via lists.openembedded.org

We can't support vgem on RHEL derived distros so disable this test for
all fedora hosts rather than specific versions.
render_hint = """If /dev/dri/renderD* is absent due to lack of suitable GPU, 'modprobe vgem' will create one suitable for mesa llvmpipe software renderer."""

I don't think this is accurate. It's the other way around: RHEL is derived from Fedora. Also, vgem module can be installed in Fedora, but not on RHEL or rocky/alma (which are derived from RHEL). We do this on all fedora build hosts, and virgl tests run fine on Fedora workers then, at least on the master branch.

What prompted this patch specifically?

Alex

#6 lore raw

Re: [OE-core][kirkstone 2/2] selftest/runtime_test/virgl: Disable for all fedora

In reply to Steve Sakoman
Quoted 14 lines · Steve Sakoman via lists.openembedded.org

We can't support vgem on RHEL derived distros so disable this test for all fedora hosts rather than specific versions.

---
 meta/lib/oeqa/selftest/cases/runtime_test.py | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)
meta/lib/oeqa/selftest/cases/runtime_test.py +0 -0
index 7dcdfd0ab2..93c6da1889 100644

meta/lib/oeqa/selftest/cases/runtime_test.py class TestImage(OESelftestTestCase): · line 261

Quoted 5 lines
meta/lib/oeqa/selftest/cases/runtime_test.py +1 -1
@@ -258,7 +258,7 @@ class TestImage(OESelftestTestCase):
 
         distro = oe.lsb.distro_identifier()
         if distro and (distro in ['debian-9', 'debian-10', 'centos-7', 'centos-8', 'ubuntu-16.04', 'ubuntu-18.04'] or
-            distro.startswith('almalinux') or distro.startswith('rocky')):
+            distro.startswith('almalinux') or distro.startswith('fedora') or distro.startswith('rocky')):

Wait, was fedora confused with centos maybe? Centos 9 is not on the list, but perhaps should be.

Alex

#4 lore raw

Re: [OE-core][kirkstone 2/2] selftest/runtime_test/virgl: Disable for all fedora

In reply to Alexander Kanavin
Quoted 8 lines · Alexander Kanavin
Quoted 3 lines · Steve Sakoman via lists.openembedded.org

We can't support vgem on RHEL derived distros so disable this test for
all fedora hosts rather than specific versions.
render_hint = """If /dev/dri/renderD* is absent due to lack of suitable GPU, 'modprobe vgem' will create one suitable for mesa llvmpipe software renderer."""

Quoted 7 lines

I don't think this is accurate. It's the other way around: RHEL is derived from Fedora. Also, vgem module can be installed in Fedora, but not on RHEL or rocky/alma (which are derived from RHEL). We do this on all fedora build hosts, and virgl tests run fine on Fedora workers then, at least on the master branch.

What prompted this patch specifically?

When we moved to the new autobuilder cluster the kirkstone supported versions of fedora were no longer available, only fedora 39/40.

The kirkstone virgl headless tests failed 100% of the time with the new fedora 39/40 workers. I discussed this with Richard on irc and he recommended that we disable the test for kirkstone on those workers.

Steve

#5 lore raw

Re: [OE-core][kirkstone 2/2] selftest/runtime_test/virgl: Disable for all fedora

Quoted 6 lines · Steve Sakoman

When we moved to the new autobuilder cluster the kirkstone supported versions of fedora were no longer available, only fedora 39/40.

The kirkstone virgl headless tests failed 100% of the time with the new fedora 39/40 workers. I discussed this with Richard on irc and he recommended that we disable the test for kirkstone on those workers.

Right, I'm ok with that then.

Alex