On 22/8/24 16:06, Marcin Juszkiewicz wrote:
On 22.08.2024 13:41, Philippe Mathieu-Daudé wrote:
# Timeouts for individual tests that can be slow e.g. with debugging
enabled
test_timeouts = {
+ 'aarch64_sbsaref' : 180,
What kind of machine is able to run those tests in 180s? I bumped them
to 2400s and got timeout (Macbook with M1 Pro).
Indeed with Avocado this timeout was per test function, now it is
per test class. Having 11 test functions, for a 1-1 change we'd
need 180 * 11 = 1980 seconds.
I made a comment on Thomas/Daniel's series this patch is based on:
https://lore.kernel.org/qemu-devel/4b4018c6-4a2b-4250-bb53-be9cc5df7...@linaro.org/
I could run all the tests in <300sec but for safety, should I keep
Avocado equivalent and use 2000s?
▶ 1/5
test_aarch64_sbsaref.Aarch64SbsarefMachine.test_sbsaref_alpine_linux_max
OK
▶ 1/5
test_aarch64_sbsaref.Aarch64SbsarefMachine.test_sbsaref_alpine_linux_max_pauth_impdef
OK
▶ 1/5
test_aarch64_sbsaref.Aarch64SbsarefMachine.test_sbsaref_alpine_linux_max_pauth_off
OK
▶ 1/5
test_aarch64_sbsaref.Aarch64SbsarefMachine.test_sbsaref_alpine_linux_neoverse_n1
OK
▶ 1/5
test_aarch64_sbsaref.Aarch64SbsarefMachine.test_sbsaref_edk2_firmware
OK
▶ 1/5
test_aarch64_sbsaref.Aarch64SbsarefMachine.test_sbsaref_openbsd73_cortex_a57
OK
▶ 1/5
test_aarch64_sbsaref.Aarch64SbsarefMachine.test_sbsaref_openbsd73_max
OK
▶ 1/5
test_aarch64_sbsaref.Aarch64SbsarefMachine.test_sbsaref_openbsd73_max_pauth_impdef
OK
▶ 1/5
test_aarch64_sbsaref.Aarch64SbsarefMachine.test_sbsaref_openbsd73_max_pauth_off
OK
▶ 1/5
test_aarch64_sbsaref.Aarch64SbsarefMachine.test_sbsaref_openbsd73_neoverse_n1
OK
1/5 qemu:func-thorough+func-aarch64-thorough+thorough /
func-aarch64-aarch64_sbsaref OK 241.79s 11 subtests
passed
"make check-avocado" (with some AVOCADO_* vars to limit list of tests)
shown me which test is run and pass/fail for each.
"make check-functional-aarch64 V=1" shows me "1/4
qemu:func-thorough+func-aarch64-thorough+thorough /
func-aarch64-aarch64_sbsaref" and timeouts without information which
tests pass, which fail.
Maybe for QEMU project this is a progress. For me it is moving tests
from working ones to "sorry, timeout, find out why" ones.