Community CI Meeting Minutes - June 12, 2025

Community CI Meeting Minutes - June 12, 2025

[Patrick Robb]

[-- Attachment #1: Type: text/plain, Size: 6371 bytes --]

#####################################################################
June 12, 2025
Attendees
1. Patrick Robb
2. Paul Szczepanek
3. Aaron Conole
4. Manit Mahajan
5. Dean Marx

#####################################################################
Minutes

=====================================================================
General Announcements
* DPDK 25.07 RC1 has been released. RC2 is expected in two weeks.
* We should consider setting a new regular timeslot(s) for this meeting. To
that end, participants please fill out this “when to meet” survey:
https://www.when2meet.com/?30965992-3ijyI
* Will we have CI Community representation at DPDK Summit Santa Clara in
September?
   * Yes, Patrick will most likely be able to go
   * There may be limited tech board representation there
   * If it were possible to do a demo demonstrating DTS running with a
limited hardware topology (even single dual interface NIC connected with
single link loopback between the two ports) that would be interesting for
participants.

=====================================================================
CI Status

---------------------------------------------------------------------
UNH-IOL Community Lab
* Manit’s new “test system environment info updater” tool which updates
test system info (like gcc version, kernel version, etc.) has been running
for a few weeks mostly without issue.
   * David did find one issue, which is that it broke the Arch Linux badge
(shows current arch compile status) for Patchwork, since the arch
environment had some info updated which broke how the Community Lab
Dashboard accessed the arch info. This is now fixed:
https://lab.dpdk.org/badges/archlinux
* Thomas raised the subject of testing wait times on Slack. There were a
couple of blocked test jobs yesterday which contributed to this but we are
mostly caught up now.
   * Also, we failed to grab a particular patchseries off Patchwork
yesterday, and this is because the latest series that had come in had a
very old series id (UNH guys were unaware that this behavior was possible).
We will need to adjust our series polling script.
   * It may make sense to make series polling process more common across
labs. UNH lab uses a database to store the most recent patch processed, and
poll-pw.sh stores the most recent patch timestamp in a file, so any new
polling script should allow the series filter timestamp to be passed in as
a script argument.
* Windows MSVC Build:
   * This test has been failing on mainline, and the compile results table
we are emailing out properly represents this, but the topline test status
for the test report is WARN instead of FAIL which is incorrect. UNH guys
will resolve ASAP.

---------------------------------------------------------------------
Intel Lab
* None

---------------------------------------------------------------------
Github Actions Robot
* Discussion at tech board about using the github robot infra to
incorporate creating PRs against a DPDK repo for an AI tool (sourcery.ai)
which will provide summaries of patchseries.
   * It would be nice if this could be pulled from Github and pushed to the
mailing list so people don’t have to look at two places.

---------------------------------------------------------------------
AWS Lab
* There have been some patches which got a meson fast tests failure
reported from AWS, though the logs in the test report body indicate all the
tests passed or skipped.
   * Shai emailed back saying it was due to an apply failure for the
release notes .rst file, and that he has seen this a couple of times
previously from other labs too and asked about us adopting a strategy to
mitigate this:
https://mails.dpdk.org/archives/test-report/2025-May/881787.html
   * Also, regardless of us adopting a strategy to mitigate this, they
should report an apply failure and bail out of running testing when this
happens
* Do we have any strategies for mitigating apply errors coming from patches
which update the release notes .rst file? No, we don’t have any strategies
for this right now other than in dpdk-ci create_series_artifact.py we try
to apply on multiple trees before reporting an apply failure.
   * Users can submit a recheck request and specify the rebase={branch}
argument to manually fix this situation if the issue is that their patch
was applied to the wrong tree.

---------------------------------------------------------------------
Loongson Lab
* Patchseries recheck support: Loongson lab is supporting the recheck
framework. Their default behavior to re-apply commits on HEAD when a
recheck is requested.
   * Patrick and Min Zhou are discussing support for the branch rebase
argument. To that end, there is a patch submitted for dpdk-ci which allows
for capturing the branch rebase argument:
https://patchwork.dpdk.org/project/ci/patch/20250611205849.72165-1-probb@iol.unh.edu/

=====================================================================
DTS Improvements & Test Development
* Virtual Function testing:
   * v3 is refactored such that topology.py handles creating the VFs on
sut_node as needed, setting trust mode on the VFs, instantiating the VF
Ports in DTS, and assigning those VF Port objects to SUT “ingress” and SUT
“egress” attribute on topology.py
      * Remaining work:
         * Add a teardown process (just echoing 0 into the sriov_numvfs
file)
         * Testsuite which sends packets between VFs
* Bugs:
   * Dean is currently investigating a bug coming from a recent patch in
which dpdk-devbind.py is not being copied to the SUT system in the correct
order, causing intermittent failures to bind the SUT ports.
   * Improve port handling patch: For one XL710 test system, no ports are
created and DTS fails during testrun setup with this patch applied.
* DTS orientation videos:
   * Dean has been filming some DTS orientation videos showing:
      * 1. How to run DTS
      * 2. Explaining the testsuite structure and how to write testsuites
      * 3. More topics to come…
* The TREX Support and nic_single_core_perf testsuite needs the v2 version
submitted by next week at the latest.

=====================================================================
Any other business
* Next Meeting Jun 12, 2025

[-- Attachment #2: Type: text/html, Size: 6878 bytes --]