Skip to contentSkip to conversation

ci: enable unit tests for non-aarch64 platforms

RFC In review Aaron Conole 6 participants 19 messages Mar 29, 2019 – Apr 17, 2019 Arm/Amlogic Meson SoC support

lore
Share & tools
lore
#0 lore raw

cover ci: enable unit tests for non-aarch64 platforms

This series is submitted as an RFC because a number of the unit tests are not successful in the travis environment. If all of them were passing, this would be submitted as PATCH instead. It could be accepted as-is but I would prefer to see all the tests passing first.

The first patch fixes up the tests to auto-detect the number of cores on a machine. This helps on lower-end systems (such as i3 laptops or something) where someone wants to verify the functionality. The number of available cores on the running system will be picked based on the running system parameters.

The second patch moves some tests out - these tests don't produce output or complete in any reasonable amount of time (10m+ for a single unit test is a little strange - they should be investigated to see if the run time can be reduced). I prefer to see these separated out since travis will completely bail if the test takes longer than 10m to produce output.

The third actually enables the testing, and runs each test leg independently. This version populates the hugepages mapping. However, it might be useful to have the option of running without hugepages enabled (and I have a separate series that can do this). However, the --no-huge flag seems to cause most of the unit tests to break since they either spawn a new instance of the EAL without passing the hugepage flags, or check against the hugepage API and use that to determine whether memory can be allocated.

Aaron Conole (3):
test/meson: auto detect number of cores
meson-tests: separate slower tests
ci: enable tests on non-arm platforms

 .ci/linux-build.sh   |  7 +++++++
 .ci/linux-setup.sh   |  6 +++++-
 app/test/meson.build | 43 +++++++++++++++++++++++++++++++++----------
 3 files changed, 45 insertions(+), 11 deletions(-)
Signature · 1 line
2.19.1
#1 lore raw

RFC 1/3 test/meson: auto detect number of cores

Some environments do not provide a minimum 4 cores for running tests. This allows those environments to still execute 'ninja test' without causing multiple failures.

---
 app/test/meson.build | 11 ++++++++++-
 1 file changed, 10 insertions(+), 1 deletion(-)
app/test/meson.build +10 -1
index ddb4d09ae..975b38daa 100644
@@ -337,10 +337,15 @@ if get_option('tests')
 	timeout_seconds = 600
 	timeout_seconds_fast = 10
 
+	# Retreive the number of CPU cores
+	num_cores = run_command('lscpu', '-p=cpu').stdout().strip().split('\n')[-1]
+	num_cores_arg = '-l 0-' + num_cores
+
+	test_args = [num_cores_arg, '-n 4']
 	foreach arg : fast_parallel_test_names
 		test(arg, dpdk_test,
 			env : ['DPDK_TEST=' + arg],
-			args : ['-c f','-n 4', '--file-prefix=@0@'.format(arg)],
+			args : test_args + ['--file-prefix=@0@'.format(arg)],
 			timeout : timeout_seconds_fast,
 			suite : 'fast-tests')
 	endforeach
@@ -348,6 +353,7 @@ if get_option('tests')
 	foreach arg : fast_non_parallel_test_names
 		test(arg, dpdk_test,
 			env : ['DPDK_TEST=' + arg],
+			args : test_args + ['--file-prefix=@0@'.format(arg)],
 			timeout : timeout_seconds_fast,
 			is_parallel : false,
 			suite : 'fast-tests')
@@ -356,6 +362,7 @@ if get_option('tests')
 	foreach arg : perf_test_names
 		test(arg, dpdk_test,
 		env : ['DPDK_TEST=' + arg],
+		args : test_args + ['--file-prefix=@0@'.format(arg)],
 		timeout : timeout_seconds,
 		is_parallel : false,
 		suite : 'perf-tests')
@@ -364,6 +371,7 @@ if get_option('tests')
 	foreach arg : driver_test_names
 		test(arg, dpdk_test,
 			env : ['DPDK_TEST=' + arg],
+			args : test_args + ['--file-prefix=@0@'.format(arg)],
 			timeout : timeout_seconds,
 			is_parallel : false,
 			suite : 'driver-tests')
@@ -372,6 +380,7 @@ if get_option('tests')
 	foreach arg : dump_test_names
 		test(arg, dpdk_test,
 			env : ['DPDK_TEST=' + arg],
+			args : test_args + ['--file-prefix=@0@'.format(arg)],
 			timeout : timeout_seconds,
 			is_parallel : false,
 			suite : 'debug-tests')
Signature · 1 line
2.19.1
#2 lore raw

Re: [RFC 1/3] test/meson: auto detect number of cores

Quoted 19 lines · Aaron Conole

Some environments do not provide a minimum 4 cores for running tests. This allows those environments to still execute 'ninja test' without causing multiple failures.

---
 app/test/meson.build | 11 ++++++++++-
 1 file changed, 10 insertions(+), 1 deletion(-)
app/test/meson.build +2 -0
index ddb4d09ae..975b38daa 100644
@@ -337,10 +337,15 @@ if get_option('tests')
        timeout_seconds = 600
        timeout_seconds_fast = 10
 
+       # Retreive the number of CPU cores
+       num_cores = run_command('lscpu',
Quoted 3 lines

'-p=cpu').stdout().strip().split('\n')[-1]
+ num_cores_arg = '-l 0-' + num_cores

Seeing how we can reduce the core number, we can at least cap it to 4 and not use all the cores on bigger systems.
But, thinking again, do you know if there is a need for more than 2 cores in the existing tests ?

+

Quoted 12 lines
app/test/meson.build +1 -1
+       test_args = [num_cores_arg, '-n 4']
        foreach arg : fast_parallel_test_names
                test(arg, dpdk_test,
                        env : ['DPDK_TEST=' + arg],
-                       args : ['-c f','-n 4', '--file-prefix=@0@

'.format(arg)],
+ args : test_args + ['--file-prefix=@0@
'.format(arg)],
timeout : timeout_seconds_fast, suite : 'fast-tests')
endforeach

app/test/meson.build if get_option('tests') · line 358

Quoted 6 lines
app/test/meson.build +1 -0
@@ -348,6 +353,7 @@ if get_option('tests')
        foreach arg : fast_non_parallel_test_names
                test(arg, dpdk_test,
                        env : ['DPDK_TEST=' + arg],
+                       args : test_args + ['--file-prefix=@0@
 '.format(arg)],
 

I don't understand this part.

(How|) was it working before ?
Is there some default args ? I could not find it.

timeout : timeout_seconds_fast,

Quoted 3 lines
app/test/meson.build +0 -0
                        is_parallel : false,

suite : 'fast-tests')

Signature · 1 line
David Marchand
#3 lore raw

Re: [RFC 1/3] test/meson: auto detect number of cores

David Marchand <[email protected]> writes:

Thanks for the review, David!

Quoted 19 lines · Aaron Conole

Some environments do not provide a minimum 4 cores for running tests. This allows those environments to still execute 'ninja test' without causing multiple failures.

---

  app/test/meson.build | 11 ++++++++++-
  1 file changed, 10 insertions(+), 1 deletion(-)

diff --git a/app/test/meson.build b/app/test/meson.build index ddb4d09ae..975b38daa 100644
--- a/app/test/meson.build
+++ b/app/test/meson.build
@@ -337,10 +337,15 @@ if get_option('tests')
timeout_seconds = 600 timeout_seconds_fast = 10

Quoted 7 lines
 +       # Retreive the number of CPU cores
 +       num_cores = run_command('lscpu', '-p=cpu').stdout().strip().split('\n')[-1]
 +       num_cores_arg = '-l 0-' + num_cores

Seeing how we can reduce the core number, we can at least cap it to 4 and not use all the cores on bigger systems.
But, thinking again, do you know if there is a need for more than 2 cores in the existing tests ?

Probably not. At least, the systems in the travis environment only provide 2 cores and most of the tests pass. OTOH, some still fail and I need to investigate them a bit more. Some of the test case failures are ex. eal flags failures and maybe they do pass some '-c f' as part of the test case, which would fail on systems without 4 cores.

Quoted 2 lines

+
+ test_args = [num_cores_arg, '-n 4']

Whoops! I think, this '-n 4' option can probably be removed as well. I think the memory channel config is optional, and I believe there's no need to try and tune it.

Quoted 14 lines

foreach arg : fast_parallel_test_names
test(arg, dpdk_test,
env : ['DPDK_TEST=' + arg],
- args : ['-c f','-n 4', '--file-prefix=@0@'.format(arg)],
+ args : test_args + ['--file-prefix=@0@'.format(arg)],
timeout : timeout_seconds_fast, suite : 'fast-tests')
endforeach
@@ -348,6 +353,7 @@ if get_option('tests')
foreach arg : fast_non_parallel_test_names
test(arg, dpdk_test,
env : ['DPDK_TEST=' + arg],
+ args : test_args + ['--file-prefix=@0@'.format(arg)],

Quoted 3 lines

I don't understand this part.

(How|) was it working before ?

It did work before.

Quoted 1 line

Is there some default args ? I could not find it.

I think it does try to do some auto-detection. I did extend this because it seemed to make sense, and I have some future work to allow passing '--no-huge' (in case we have a CI system that doesn't allow allocating hugepages). Maybe it doesn't make sense to add that right now, though. I'll try without it.

Quoted 3 lines

timeout : timeout_seconds_fast, is_parallel : false, suite : 'fast-tests')

#4 lore raw

Re: [RFC 1/3] test/meson: auto detect number of cores

In reply to Aaron Conole
Quoted 8 lines

-----Original Message-----
From: dev [mailto:[email protected]] On Behalf Of Aaron Conole
Sent: Friday, March 29, 2019 5:23 PM
To: [email protected]
Subject: [dpdk-dev] [RFC 1/3] test/meson: auto detect number of cores

Quoted 4 lines

Some environments do not provide a minimum 4 cores for running tests. This allows those environments to still execute 'ninja test' without causing multiple failures.

FYI, there is a patch http://patches.dpdk.org/patch/50850/ doing something similar.

--file-prefix is not supported on FreeBSD, so how about you also include the related fix from above patch link into your patch series?

Thanks,
Reshma

#5 lore raw

Re: [RFC 1/3] test/meson: auto detect number of cores

Quoted 11 lines · Reshma
Quoted 10 lines

-----Original Message-----
From: dev [mailto:[email protected]] On Behalf Of Aaron Conole
Sent: Friday, March 29, 2019 5:23 PM
To: [email protected]
Subject: [dpdk-dev] [RFC 1/3] test/meson: auto detect number of cores

Some environments do not provide a minimum 4 cores for running tests. This allows those environments to still execute 'ninja test' without causing multiple failures.

Quoted 4 lines

FYI, there is a patch http://patches.dpdk.org/patch/50850/ doing something similar.

--file-prefix is not supported on FreeBSD, so how about you also include the related fix from above patch link into your patch series?

Is that other patch going to be accepted? Then I can drop this patch from my series. Otherwise, I'll fold in such a change with my next version.

Quoted 2 lines

Thanks,
Reshma

#6 lore raw

Re: [RFC 1/3] test/meson: auto detect number of cores

01/04/2019 19:48, Aaron Conole:

Quoted 8 lines
Quoted 7 lines · Reshma

From: dev [mailto:[email protected]] On Behalf Of Aaron Conole

Quoted 5 lines

Some environments do not provide a minimum 4 cores for running tests. This allows those environments to still execute 'ninja test' without causing multiple failures.

Quoted 8 lines
Quoted 4 lines

FYI, there is a patch http://patches.dpdk.org/patch/50850/ doing something similar.

--file-prefix is not supported on FreeBSD, so how about you also include the related fix from above patch link into your patch series?

Is that other patch going to be accepted? Then I can drop this patch from my series. Otherwise, I'll fold in such a change with my next version.

Sorry, I'm late on applying patches for unit tests.
As I am not sure which one will get the best review first, you should keep it with a note that it conflicts with the other one.

#7 lore raw

Re: [RFC 1/3] test/meson: auto detect number of cores

Quoted 14 lines · Thomas Monjalon

01/04/2019 19:48, Aaron Conole:

Quoted 13 lines
Quoted 11 lines · Reshma

From: dev [mailto:[email protected]] On Behalf Of Aaron Conole

Quoted 5 lines

Some environments do not provide a minimum 4 cores for running tests. This allows those environments to still execute 'ninja test' without causing multiple failures.

FYI, there is a patch http://patches.dpdk.org/patch/50850/ doing something similar.

--file-prefix is not supported on FreeBSD, so how about you also include the related fix from above patch link into your patch series?

Quoted 7 lines
Quoted 3 lines

Is that other patch going to be accepted? Then I can drop this patch from my series. Otherwise, I'll fold in such a change with my next version.

Sorry, I'm late on applying patches for unit tests.
As I am not sure which one will get the best review first, you should keep it with a note that it conflicts with the other one.

Okay - thanks, Thomas!

#8 lore raw

RFC 2/3 meson-tests: separate slower tests

In reply to Aaron Conole

When running some tests in the CI environment, the run time can take in excess of 10 minutes to produce output. This causes the entire CI run not to succeed. Separate these tests so that they can be run from a normal environment, and deselected in a CI environment.

There's also a small whitespace fix bundled in for the perf_tests block.

---
 app/test/meson.build | 34 ++++++++++++++++++++++++----------
 1 file changed, 24 insertions(+), 10 deletions(-)
app/test/meson.build +24 -10
index 975b38daa..1b0ff0523 100644
@@ -171,7 +171,6 @@ fast_parallel_test_names = [
         'red_autotest',
         'ring_autotest',
         'ring_pmd_autotest',
-        'rwlock_autotest',
         'sched_autotest',
         'spinlock_autotest',
         'string_autotest',
@@ -214,15 +213,21 @@ fast_non_parallel_test_names = [
         'thash_autotest',
 ]
 
+# These test cases are non-parallel and also take > 10m to execute in
+# some environments (such as CI systems), so allow deselecting them
+slow_test_names = [
+        'rwlock_autotest',
+        'mempool_perf_autotest',
+        'reciprocal_division',
+        'reciprocal_division_perf',
+]
+
 # All test cases in perf_test_names list are non-parallel
 perf_test_names = [
         'ring_perf_autotest',
-        'mempool_perf_autotest',
         'memcpy_perf_autotest',
         'hash_perf_autotest',
         'timer_perf_autotest',
-        'reciprocal_division',
-        'reciprocal_division_perf',
         'lpm_perf_autotest',
         'red_all',
         'barrier_autotest',
@@ -332,7 +337,7 @@ if get_option('tests')
 		install_rpath: driver_install_path,
 		install: false)
 
-	# some perf tests (eg: memcpy perf autotest)take very long
+	# some tests (eg: the slow tests) take very long
 	# to complete, so timeout to 10 minutes
 	timeout_seconds = 600
 	timeout_seconds_fast = 10
@@ -359,13 +364,22 @@ if get_option('tests')
 			suite : 'fast-tests')
 	endforeach
 
+	foreach arg : slow_test_names
+		test(arg, dpdk_test,
+			env : ['DPDK_TEST=' + arg],
+			args : test_args + ['--file-prefix=@0@'.format(arg)],
+			timeout : timeout_seconds,
+			is_parallel : false,
+			suite : 'slow-tests')
+	endforeach
+
 	foreach arg : perf_test_names
 		test(arg, dpdk_test,
-		env : ['DPDK_TEST=' + arg],
-		args : test_args + ['--file-prefix=@0@'.format(arg)],
-		timeout : timeout_seconds,
-		is_parallel : false,
-		suite : 'perf-tests')
+			env : ['DPDK_TEST=' + arg],
+			args : test_args + ['--file-prefix=@0@'.format(arg)],
+			timeout : timeout_seconds,
+			is_parallel : false,
+			suite : 'perf-tests')
 	endforeach
 
 	foreach arg : driver_test_names
Signature · 1 line
2.19.1
#9 lore raw

RFC 3/3 ci: enable tests on non-arm platforms

In reply to Aaron Conole

When building under Travis (or another linux CI service), enable running the tests for each build.

---
 .ci/linux-build.sh | 7 +++++++
 .ci/linux-setup.sh | 6 +++++-
 2 files changed, 12 insertions(+), 1 deletion(-)
.ci/linux-build.sh +7 -0
index 4eb7c3cf0..72510a433 100755
@@ -22,3 +22,10 @@ fi
 OPTS="$OPTS --default-library=$DEF_LIB"
 meson build --werror -Dexamples=all $OPTS
 ninja -C build
+
+if [ "$AARCH64" != "1" ]; then
+    sudo meson test -C build --suite fast-tests
+    sudo meson test -C build --suite perf-tests
+    sudo meson test -C build --suite driver-tests
+    sudo meson test -C build --suite debug-tests
+fi
.ci/linux-setup.sh +5 -1
index acdf9f370..18d2b22fc 100755
@@ -1,3 +1,7 @@
 #!/bin/sh
 
-python3 -m pip install --upgrade meson --user
+sudo python3 -m pip install --upgrade meson
+
+cat /proc/meminfo
+sudo sh -c 'echo 1024 > /proc/sys/vm/nr_hugepages'
+cat /proc/meminfo
Signature · 1 line
2.19.1
#10 lore raw

Re: [dpdk-dev] [RFC 3/3] ci: enable tests on non-arm platforms

Quoted 20 lines

-----Original Message-----
From: dev <[email protected]> On Behalf Of Aaron Conole
Sent: Friday, March 29, 2019 10:53 PM
To: [email protected]
Subject: [dpdk-dev] [RFC 3/3] ci: enable tests on non-arm platforms

When building under Travis (or another linux CI service), enable running the tests for each build.

---
 .ci/linux-build.sh | 7 +++++++
 .ci/linux-setup.sh | 6 +++++-
 2 files changed, 12 insertions(+), 1 deletion(-)
.ci/linux-build.sh index 4eb7c3cf0..72510a433 renamed from .ci/linux-build.sh +0 -0

100755

.ci/linux-build.sh fi · line 25

Quoted 4 lines
.ci/linux-build.sh +2 -0
@@ -22,3 +22,10 @@ fi
 OPTS="$OPTS --default-library=$DEF_LIB"
 meson build --werror -Dexamples=all $OPTS  ninja -C build
+
+if [ "$AARCH64" != "1" ]; then

If the indention is run only on X86 then, X86 == 1 would be the correct logic as we do have powerpc and arm32 support in dpdk.

Any reason why we are opting out arm64?
There are some cloud hosting services with arm64 as well.

https://www.packet.com/cloud/servers/c1-large-arm/

#11 lore raw

Re: [dpdk-dev] [RFC 3/3] ci: enable tests on non-arm platforms

Quoted 15 lines · Jerin Jacob Kollanukkaran
Quoted 15 lines

-----Original Message-----
From: dev <[email protected]> On Behalf Of Aaron Conole
Sent: Friday, March 29, 2019 10:53 PM
To: [email protected]
Subject: [dpdk-dev] [RFC 3/3] ci: enable tests on non-arm platforms

When building under Travis (or another linux CI service), enable running the tests for each build.

---
 .ci/linux-build.sh | 7 +++++++
 .ci/linux-setup.sh | 6 +++++-
 2 files changed, 12 insertions(+), 1 deletion(-)
Quoted 12 lines

.ci/linux-build.sh fi · line 25

Quoted 9 lines
.ci/linux-build.sh index 4eb7c3cf0..72510a433 renamed from .ci/linux-build.sh +0 -0

100755

.ci/linux-build.sh +2 -0
@@ -22,3 +22,10 @@ fi
 OPTS="$OPTS --default-library=$DEF_LIB"
 meson build --werror -Dexamples=all $OPTS  ninja -C build
+
+if [ "$AARCH64" != "1" ]; then

If the indention is run only on X86 then, X86 == 1 would be the correct logic as we do have powerpc and arm32 support in dpdk.

Correct - I should have used a more robust and targeted test.

Quoted 4 lines

Any reason why we are opting out arm64?
There are some cloud hosting services with arm64 as well.

https://www.packet.com/cloud/servers/c1-large-arm/

There is At the moment, we aren't hooked up to any - do you know of one that we can easily integrate with right now? AFAICT, there are hosting services, but we would still need all the other things that go with CI service (dashboards, scheduling, tracking, etc.) and it's a big project to set up, maintain, etc.

#12 lore raw

Re: [dpdk-dev] [EXT] Re: [RFC 3/3] ci: enable tests on non-arm platforms

Quoted 10 lines

-----Original Message-----
From: Aaron Conole <[email protected]>
Sent: Wednesday, April 17, 2019 6:10 PM
To: Jerin Jacob Kollanukkaran <[email protected]>
Cc: [email protected]
Subject: [EXT] Re: [dpdk-dev] [RFC 3/3] ci: enable tests on non-arm platforms

----------------------------------------------------------------------

Quoted 33 lines
Quoted 15 lines · Jerin Jacob Kollanukkaran
Quoted 15 lines

-----Original Message-----
From: dev <[email protected]> On Behalf Of Aaron Conole
Sent: Friday, March 29, 2019 10:53 PM
To: [email protected]
Subject: [dpdk-dev] [RFC 3/3] ci: enable tests on non-arm platforms

When building under Travis (or another linux CI service), enable running the tests for each build.

---
 .ci/linux-build.sh | 7 +++++++
 .ci/linux-setup.sh | 6 +++++-
 2 files changed, 12 insertions(+), 1 deletion(-)
Quoted 13 lines

.ci/linux-build.sh fi · line 25

Quoted 10 lines
.ci/linux-build.sh index renamed from .ci/linux-build.sh +0 -0

4eb7c3cf0..72510a433
100755

.ci/linux-build.sh +2 -0
@@ -22,3 +22,10 @@ fi
 OPTS="$OPTS --default-library=$DEF_LIB"
 meson build --werror -Dexamples=all $OPTS  ninja -C build
+
+if [ "$AARCH64" != "1" ]; then

If the indention is run only on X86 then, X86 == 1 would be the correct logic as we do have powerpc and arm32 support in dpdk.

Correct - I should have used a more robust and targeted test.

Quoted 7 lines
Quoted 4 lines

Any reason why we are opting out arm64?
There are some cloud hosting services with arm64 as well.

https://www.packet.com/cloud/servers/c1-large-arm/

There is At the moment, we aren't hooked up to any - do you know of one that we can easily integrate with right now? AFAICT, there are hosting services, but

I know, Following cloud hosting providers has arm64 support. But, not sure about the feature set for CI work.
(2) and (3) are based on ThunderX servers.

https://www.linaro.cloud/
https://www.packet.com/cloud/servers/c1-large-arm/
www.scaleway.com

Quoted 2 lines

we would still need all the other things that go with CI service (dashboards, scheduling, tracking, etc.) and it's a big project to set up, maintain, etc.

#13 lore raw
In reply to Aaron Conole
Quoted 35 lines · Aaron Conole

This series is submitted as an RFC because a number of the unit tests are not successful in the travis environment. If all of them were passing, this would be submitted as PATCH instead. It could be accepted as-is but I would prefer to see all the tests passing first.

The first patch fixes up the tests to auto-detect the number of cores on a machine. This helps on lower-end systems (such as i3 laptops or something)
where someone wants to verify the functionality. The number of available cores on the running system will be picked based on the running system parameters.

The second patch moves some tests out - these tests don't produce output or complete in any reasonable amount of time (10m+ for a single unit test is a little strange - they should be investigated to see if the run time can be reduced). I prefer to see these separated out since travis will completely
bail if the test takes longer than 10m to produce output.

The third actually enables the testing, and runs each test leg independently.
This version populates the hugepages mapping. However, it might be useful to have the option of running without hugepages enabled (and I have a separate series that can do this). However, the --no-huge flag seems to cause
most of the unit tests to break since they either spawn a new instance of the EAL without passing the hugepage flags, or check against the hugepage API
and use that to determine whether memory can be allocated.

Aaron Conole (3):
test/meson: auto detect number of cores
meson-tests: separate slower tests
ci: enable tests on non-arm platforms

Quoted 5 lines
 .ci/linux-build.sh   |  7 +++++++
 .ci/linux-setup.sh   |  6 +++++-
 app/test/meson.build | 43 +++++++++++++++++++++++++++++++++----------
 3 files changed, 45 insertions(+), 11 deletions(-)

I tried using meson/ninja for the tests, something that bothered me is that I can't interrupt the tests.
I had to kill manually, meson, ninja and I had some leftover dpdk-test processes (maybe due to some ^Z I hit...).
Is this expected ?

This is quite frustrating when testing "before" and "after" each patch.

Signature · 1 line
David Marchand
#14 lore raw
Quoted 37 lines · David Marchand

On Fri, Mar 29, 2019 at 6:23 PM Aaron Conole <[email protected]> wrote:

This series is submitted as an RFC because a number of the unit tests are not successful in the travis environment. If all of them were passing, this would be submitted as PATCH instead. It could be accepted as-is but I would prefer to see all the tests passing first.

The first patch fixes up the tests to auto-detect the number of cores on a machine. This helps on lower-end systems (such as i3 laptops or something) where someone wants to verify the functionality. The number of available cores on the running system will be picked based on the running system parameters.

The second patch moves some tests out - these tests don't produce output or complete in any reasonable amount of time (10m+ for a single unit test is a little strange - they should be investigated to see if the run time can be reduced). I prefer to see these separated out since travis will completely bail if the test takes longer than 10m to produce output.

The third actually enables the testing, and runs each test leg independently. This version populates the hugepages mapping. However, it might be useful to have the option of running without hugepages enabled (and I have a separate series that can do this). However, the --no-huge flag seems to cause most of the unit tests to break since they either spawn a new instance of the EAL without passing the hugepage flags, or check against the hugepage API and use that to determine whether memory can be allocated.

Aaron Conole (3):
test/meson: auto detect number of cores
meson-tests: separate slower tests
ci: enable tests on non-arm platforms

  .ci/linux-build.sh   |  7 +++++++
  .ci/linux-setup.sh   |  6 +++++-
  app/test/meson.build | 43 +++++++++++++++++++++++++++++++++----------
  3 files changed, 45 insertions(+), 11 deletions(-)
Quoted 4 lines

I tried using meson/ninja for the tests, something that bothered me is that I can't interrupt the tests. I had to kill manually, meson, ninja and I had some leftover dpdk-test processes (maybe due to some ^Z I hit...).
Is this expected ?

Certainly not by me. I usually let everything complete, though (which takes a looong time if I run the full suite).

Quoted 1 line

This is quite frustrating when testing "before" and "after" each patch.

Agreed. :-/

I'll have to try it out to see what's happening. Does it only happen with this series? I'd be surprised, but possibly I introduced some error.

#15 lore raw
Quoted 11 lines · Aaron Conole
Quoted 1 line · David Marchand

I tried using meson/ninja for the tests, something that bothered me is

that I can't interrupt the tests.

Quoted 1 line

I had to kill manually, meson, ninja and I had some leftover dpdk-test

processes (maybe due to some ^Z I

Quoted 2 lines

hit...).
Is this expected ?

Certainly not by me. I usually let everything complete, though (which takes a looong time if I run the full suite).

Quoted 7 lines
Quoted 1 line

This is quite frustrating when testing "before" and "after" each patch.

Agreed. :-/

I'll have to try it out to see what's happening. Does it only happen with this series? I'd be surprised, but possibly I introduced some error.

Nop, I got this even before your first patch.

Signature · 1 line
David Marchand
#16 lore raw
Quoted 15 lines · David Marchand
Quoted 13 lines · Aaron Conole
Quoted 1 line · David Marchand

I tried using meson/ninja for the tests, something that bothered me is

that I can't interrupt the tests.

Quoted 1 line

I had to kill manually, meson, ninja and I had some leftover dpdk-test

processes (maybe due to some ^Z I

Quoted 2 lines

hit...).
Is this expected ?

Certainly not by me. I usually let everything complete, though (which takes a looong time if I run the full suite).

Quoted 1 line

This is quite frustrating when testing "before" and "after" each patch.

Quoted 8 lines
Quoted 5 lines

Agreed. :-/

I'll have to try it out to see what's happening. Does it only happen with this series? I'd be surprised, but possibly I introduced some error.

Nop, I got this even before your first patch.

Is this meson related or related to the auto test binary in DPDK. I know traditionally I've found the test binary rather difficult to kill, but I'd like to be sure that the meson infrastructure itself isn't making it worse.

/Bruce

#17 lore raw
Quoted 26 lines · Bruce Richardson
Quoted 4 lines · David Marchand
Quoted 2 lines · Aaron Conole
Quoted 1 line · David Marchand

I tried using meson/ninja for the tests, something that bothered me

is

Quoted 2 lines
Quoted 2 lines

that I can't interrupt the tests.

Quoted 1 line

I had to kill manually, meson, ninja and I had some leftover

dpdk-test

Quoted 8 lines
Quoted 8 lines

processes (maybe due to some ^Z I

Quoted 2 lines

hit...).
Is this expected ?

Certainly not by me. I usually let everything complete, though (which takes a looong time if I run the full suite).

Quoted 1 line

This is quite frustrating when testing "before" and "after" each

patch.

Quoted 5 lines
Quoted 5 lines

Agreed. :-/

I'll have to try it out to see what's happening. Does it only happen with this series? I'd be surprised, but possibly I introduced some

error.

Quoted 2 lines
Quoted 1 line
Quoted 7 lines
Quoted 2 lines

Nop, I got this even before your first patch.

Is this meson related or related to the auto test binary in DPDK. I know traditionally I've found the test binary rather difficult to kill, but I'd like to be sure that the meson infrastructure itself isn't making it worse.

Hard to tell, I would have to retest and investigate, unless Aaron went further than me.

Signature · 1 line
David Marchand
#18 lore raw
Quoted 24 lines · David Marchand

On Tue, Apr 2, 2019 at 11:37 AM Bruce Richardson <[email protected]> wrote:

Quoted 21 lines · David Marchand
Quoted 18 lines · Aaron Conole
Quoted 1 line · David Marchand

I tried using meson/ninja for the tests, something that bothered me is

that I can't interrupt the tests.

Quoted 1 line

I had to kill manually, meson, ninja and I had some leftover dpdk-test

processes (maybe due to some ^Z I

Quoted 2 lines

hit...).
Is this expected ?

Certainly not by me. I usually let everything complete, though (which takes a looong time if I run the full suite).

Quoted 1 line

This is quite frustrating when testing "before" and "after" each patch.

Agreed. :-/

I'll have to try it out to see what's happening. Does it only happen with this series? I'd be surprised, but possibly I introduced some error.

Quoted 8 lines
Quoted 2 lines

Nop, I got this even before your first patch.

Is this meson related or related to the auto test binary in DPDK. I know traditionally I've found the test binary rather difficult to kill, but I'd like to be sure that the meson infrastructure itself isn't making it worse.

Hard to tell, I would have to retest and investigate, unless Aaron went further than me.

I did some investigation with this. At least I don't see any lingering process from meson, but it does take time for the tests to die when I hit CTRL-C, and I get a warning - I found this related bug:
https://github.com/mesonbuild/meson/issues/2281

I guess it could be some kind of interplay between the way meson kills tests? Looking at the commit that eventually 'closes' the bug, they merely set a flag rather than pass a kill signal down, so I guess it's probably never going to be immediate.