Hi Aaron/Cristian,
Ferruh Yigit <ferruh.yigit@amd.com> writes:
> On 2/13/2024 5:38 PM, Cristian Dumitrescu wrote:
>> When rte_log.h was moved to a new directory, the include path was not
>> updated for the generated C code produced by the pipeline library,
>> which results in build failure for this code.
>>
>> Fixes: 09ce41310930 ("log: separate logging functions out of EAL")
>> Cc: stable@dpdk.org
>>
>
> Hi Cristian,
>
> How can I verify the fix? Can you please list the required steps?
I guess maybe (?) with the pipeline DTS case, but I'm not sure that
would be sufficient.
> And I wonder how this skipped the testing, I guess v23.11 released with
> this defect. Is there a gap in the CI or internal build/test scripts?
I don't know that softnic driver is used in the lab. Actually, would
DTS suite even have triggered this issue? I'm not sure if there is a
set of tests which covers the case. Maybe Patrick can confirm about the
pipeline test?
So, based on what Cristian stated (passing in any of the .cli files when starting the pipeline example app would show it is fixed), yes I assume the DTS testsuite would have caught this, as I can see the testsuite does do that. But, yes it's also true that the pipeline testsuites are not run at UNH or the Intel Lab (the two labs which publicly report DTS results), so that's how this gets through CI Testing. It is not possible (testing capacity wise) to run every testsuite, and I don't think there has been conversation between the lab and our vendor contacts about this specific coverage (at least not while I've been working here).
However, based on the physical testplan requirements (4 10G tester ports to 4 10G DUT ports), we could bring the testsuite online if there is interest(and the testsuite hasn't broken since it dropped in 2020). One of our Intel testbeds which we run currently has exactly that NIC topology, and it also doesn't have bad testing capacity concerns as compared to some other testbeds. Let me know if there is an interest in this coverage and I'll make a ticket for the team to take a look. At a minimum we could dry run the framework on the testbed I'm thinking of and provide feedback, which I suppose would take only a couple minutes/hours. Christian let me know if that's a value - it's a low barrier of entry to dry run.