From: Gustavo Romero <gustavo.romero@linaro.org>
To: "Pierrick Bouvier" <pierrick.bouvier@linaro.org>,
"Alex Bennée" <alex.bennee@linaro.org>,
qemu-devel@nongnu.org
Cc: "Manos Pitsidianakis" <manos.pitsidianakis@linaro.org>,
"Philippe Mathieu-Daudé" <philmd@linaro.org>
Subject: Re: [PULL 11/14] tests/tcg: reduce the number of plugin tests combinations
Date: Thu, 31 Jul 2025 15:58:24 -0300 [thread overview]
Message-ID: <a21427c9-9b10-45fe-804f-9851f31fc69b@linaro.org> (raw)
In-Reply-To: <2635ef88-b662-41b9-b206-ca1923e87e91@linaro.org>
Hi,
On 7/28/25 17:14, Gustavo Romero wrote:
> Hi,
>
> On 7/28/25 16:03, Pierrick Bouvier wrote:
>> Hi Alex,
>>
>> On 7/27/25 1:32 AM, Alex Bennée wrote:
>>> As our set of multiarch tests has grown the practice of running every
>>> plugin with every test is becoming unsustainable. If we switch to
>>> ensuring every test gets run with at least one plugin we can speed
>>> things up.
>>>
>>> Some plugins do need to be run with specific tests (for example the
>>> memory instrumentation test). We can handle this by manually adding
>>> them to EXTRA_RUNS. We also need to wrap rules in a CONFIG_PLUGIN test
>>> so we don't enable the runs when plugins are not enabled.
>>>
>>> Reviewed-by: Manos Pitsidianakis <manos.pitsidianakis@linaro.org>
>>> Signed-off-by: Alex Bennée <alex.bennee@linaro.org>
>>> Message-ID: <20250725154517.3523095-12-alex.bennee@linaro.org>
>>>
>>> diff --git a/tests/tcg/Makefile.target b/tests/tcg/Makefile.target
>>> index a12b15637ea..18afd5be194 100644
>>> --- a/tests/tcg/Makefile.target
>>> +++ b/tests/tcg/Makefile.target
>>> @@ -173,14 +173,25 @@ PLUGINS=$(filter-out $(DISABLE_PLUGINS), \
>>> # We need to ensure expand the run-plugin-TEST-with-PLUGIN
>>> # pre-requistes manually here as we can't use stems to handle it. We
>>> # only expand MULTIARCH_TESTS which are common on most of our targets
>>> -# to avoid an exponential explosion as new tests are added. We also
>>> -# add some special helpers the run-plugin- rules can use below.
>>> +# and rotate the plugins so we don't grow too out of control as new
>>> +# tests are added. Plugins that need to run with a specific test
>>> +# should ensure they add their combination to EXTRA_RUNS.
>>> ifneq ($(MULTIARCH_TESTS),)
>>> -$(foreach p,$(PLUGINS), \
>>> - $(foreach t,$(MULTIARCH_TESTS),\
>>> - $(eval run-plugin-$(t)-with-$(p): $t $p) \
>>> - $(eval RUN_TESTS+=run-plugin-$(t)-with-$(p))))
>>> +
>>> +NUM_PLUGINS := $(words $(PLUGINS))
>>> +NUM_TESTS := $(words $(MULTIARCH_TESTS))
>>> +
>>> +define mod_plus_one
>>> + $(shell $(PYTHON) -c "print( ($(1) % $(2)) + 1 )")
>>> +endef
>>> +
>>> +$(foreach _idx, $(shell seq 1 $(NUM_TESTS)), \
>>> + $(eval _test := $(word $(_idx), $(MULTIARCH_TESTS))) \
>>> + $(eval _plugin := $(word $(call mod_plus_one, $(_idx), $(NUM_PLUGINS)), $(PLUGINS))) \
>>> + $(eval run-plugin-$(_test)-with-$(_plugin): $(_test) $(_plugin)) \
>>> + $(eval RUN_TESTS+=run-plugin-$(_test)-with-$(_plugin)))
>>> +
>>> endif # MULTIARCH_TESTS
>>> endif # CONFIG_PLUGIN
>>> diff --git a/tests/tcg/multiarch/Makefile.target b/tests/tcg/multiarch/Makefile.target
>>> index bfdf7197a7b..38345ff8805 100644
>>> --- a/tests/tcg/multiarch/Makefile.target
>>> +++ b/tests/tcg/multiarch/Makefile.target
>>> @@ -189,6 +189,10 @@ run-plugin-semiconsole-with-%:
>>> TESTS += semihosting semiconsole
>>> endif
>>> +test-plugin-mem-access: CFLAGS+=-pthread -O0
>>> +test-plugin-mem-access: LDFLAGS+=-pthread -O0
>>> +
>>> +ifeq ($(CONFIG_PLUGIN),y)
>>> # Test plugin memory access instrumentation
>>> run-plugin-test-plugin-mem-access-with-libmem.so: \
>>> PLUGIN_ARGS=$(COMMA)print-accesses=true
>>> @@ -197,8 +201,8 @@ run-plugin-test-plugin-mem-access-with-libmem.so: \
>>> $(SRC_PATH)/tests/tcg/multiarch/check-plugin-output.sh \
>>> $(QEMU) $<
>>> -test-plugin-mem-access: CFLAGS+=-pthread -O0
>>> -test-plugin-mem-access: LDFLAGS+=-pthread -O0
>>> +EXTRA_RUNS += run-plugin-test-plugin-mem-access-with-libmem.so
>>> +endif
>>> # Update TESTS
>>> TESTS += $(MULTIARCH_TESTS)
>>> diff --git a/tests/tcg/multiarch/system/Makefile.softmmu-target b/tests/tcg/multiarch/system/Makefile.softmmu-target
>>> index 5acf2700812..4171b4e6aa0 100644
>>> --- a/tests/tcg/multiarch/system/Makefile.softmmu-target
>>> +++ b/tests/tcg/multiarch/system/Makefile.softmmu-target
>>> @@ -71,8 +71,11 @@ endif
>>> MULTIARCH_RUNS += run-gdbstub-memory run-gdbstub-interrupt \
>>> run-gdbstub-untimely-packet run-gdbstub-registers
>>> +ifeq ($(CONFIG_PLUGIN),y)
>>> # Test plugin memory access instrumentation
>>> -run-plugin-memory-with-libmem.so: \
>>> - PLUGIN_ARGS=$(COMMA)region-summary=true
>>> -run-plugin-memory-with-libmem.so: \
>>> - CHECK_PLUGIN_OUTPUT_COMMAND=$(MULTIARCH_SYSTEM_SRC)/validate-memory-counts.py $@.out
>>> +run-plugin-memory-with-libmem.so: memory libmem.so
>>> +run-plugin-memory-with-libmem.so: PLUGIN_ARGS=$(COMMA)region-summary=true
>>> +run-plugin-memory-with-libmem.so: CHECK_PLUGIN_OUTPUT_COMMAND=$(MULTIARCH_SYSTEM_SRC)/validate-memory-counts.py $@.out
>>> +
>>> +EXTRA_RUNS += run-plugin-memory-with-libmem.so
>>> +endif
>>
>> I'm not sure how it's related, but check-tcg on aarch64 host now fails [1] since this series was merged, and I suspect it may be related to this patch. I didn't spend time to reproduce and investigate it.
>>
>> [1] https://github.com/pbo-linaro/qemu-ci/actions/runs/16575679153/job/46879690693
>
> I was not able to reproduce it locally on 22.04, but in the CI indeed the test command is missing the "test-plugin-mem-access" binary at the end, just before redirection to .so.out, it should be:
>
> [...] -D test-plugin-mem-access-with-libmem.so.pout test-plugin-mem-access > run-plugin-test-plugin-mem-access-with-libmem.so.out
>
> not:
>
> [...] -D test-plugin-mem-access-with-libmem.so.pout > run-plugin-test-plugin-mem-access-with-libmem.so.out
The problem is that the new rules generated by a shuffled combination of tests and plugin
(the one using eval) sometimes does not include the rule for the test+plugin combination
passed via EXTRA_RUNs.
So EXTRA_RUNS tests like run-plugin-test-plugin-mem-access-with-libmem.so might end up
with a proper rule if the test test-plugin-mem-access is combined with other plugin
randomly, since now a plugin is picked up based on a tests index _idx modulo with the
number of plugins.
A possible fix is to generate correctly the rules for the tests with plugins passed
via EXTRA_RUNS and ideally, following the patch's mood, remove the test from running
if any other plugin (no shuffle with any other plugin).
Of course, as Pierrick said, this patch could be a premature optimization. So maybe
it doesn´t justify adding more complexity to our Makefile (Makefiles are hard to debug,
with 'eval' tricks, it's _reallly_ hard, so the more we avoid it the better).
That said, if we want to keep this commit, I've kicked off a test to fix it here:
https://gitlab.com/gusbromero/qemu/-/pipelines/1959953122
and sent the fix to the ML too. HTH.
Cheers,
Gustavo
next prev parent reply other threads:[~2025-07-31 20:33 UTC|newest]
Thread overview: 28+ messages / expand[flat|nested] mbox.gz Atom feed top
2025-07-27 8:32 [PULL for 10.1-rc1 00/14] documentation an testing updates Alex Bennée
2025-07-27 8:32 ` [PULL 01/14] docs/user: clarify user-mode expects the same OS Alex Bennée
2025-07-27 8:32 ` [PULL 02/14] docs/system: reword the TAP notes to remove tarball ref Alex Bennée
2025-07-27 8:32 ` [PULL 03/14] docs/user: clean up headings Alex Bennée
2025-07-27 8:32 ` [PULL 04/14] docs/user: slightly reword section on system calls Alex Bennée
2025-07-27 8:32 ` [PULL 05/14] docs/user: expand section on threading Alex Bennée
2025-07-27 8:32 ` [PULL 06/14] tests/functional: add hypervisor test for aarch64 Alex Bennée
2025-08-13 7:27 ` Philippe Mathieu-Daudé
2025-08-13 8:56 ` Alex Bennée
2025-07-27 8:32 ` [PULL 07/14] tests/tcg: skip libsyscall.so on softmmu tests Alex Bennée
2025-07-27 8:32 ` [PULL 08/14] tests/tcg: remove ADDITIONAL_PLUGINS_TESTS Alex Bennée
2025-07-27 8:32 ` [PULL 09/14] tests/tcg: don't include multiarch tests if not supported Alex Bennée
2025-07-27 8:32 ` [PULL 10/14] configure: expose PYTHON to test/tcg/config-host.mak Alex Bennée
2025-07-27 8:32 ` [PULL 11/14] tests/tcg: reduce the number of plugin tests combinations Alex Bennée
2025-07-28 19:03 ` Pierrick Bouvier
2025-07-28 20:14 ` Gustavo Romero
2025-07-31 18:58 ` Gustavo Romero [this message]
2025-07-31 19:23 ` Pierrick Bouvier
2025-07-31 19:34 ` Gustavo Romero
2025-08-01 0:29 ` Gustavo Romero
2025-08-01 0:33 ` Gustavo Romero
2025-07-30 7:25 ` Pierrick Bouvier
2025-08-04 8:39 ` Alex Bennée
2025-08-04 16:24 ` Pierrick Bouvier
2025-07-27 8:32 ` [PULL 12/14] tests/docker: add --arch-only to qemu deps for all-test-cross Alex Bennée
2025-07-27 8:32 ` [PULL 13/14] tests/docker: handle host-arch selection " Alex Bennée
2025-07-27 8:32 ` [PULL 14/14] tests/functional: expose sys.argv to unittest.main Alex Bennée
2025-07-28 15:20 ` [PULL for 10.1-rc1 00/14] documentation an testing updates Stefan Hajnoczi
Reply instructions:
You may reply publicly to this message via plain-text email
using any one of the following methods:
* Save the following mbox file, import it into your mail client,
and reply-to-all from there: mbox
Avoid top-posting and favor interleaved quoting:
https://en.wikipedia.org/wiki/Posting_style#Interleaved_style
* Reply using the --to, --cc, and --in-reply-to
switches of git-send-email(1):
git send-email \
--in-reply-to=a21427c9-9b10-45fe-804f-9851f31fc69b@linaro.org \
--to=gustavo.romero@linaro.org \
--cc=alex.bennee@linaro.org \
--cc=manos.pitsidianakis@linaro.org \
--cc=philmd@linaro.org \
--cc=pierrick.bouvier@linaro.org \
--cc=qemu-devel@nongnu.org \
/path/to/YOUR_REPLY
https://kernel.org/pub/software/scm/git/docs/git-send-email.html
* If your mail client supports setting the In-Reply-To header
via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line
before the message body.
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for NNTP newsgroup(s).