#####################################################################
Attendees
* Patrick Robb
* Jeremy Spewock
* Honnappa Nagarahalli
* Luca Vizzarro
* Nicholas Pratte
#####################################################################
Minutes
=====================================================================
General Announcements
* DPDK 24.03 will release this week (releasing today, last I knew)
* Lab will be sending a few patches which we have reviewed
internally, for the 24.07 release
* UNH Lab will have people at DPDK Summit (likely to be Montreal). So
we can give lab updates and/or DTS updates, or anything else the group
thinks is relevant
* Honnappa notes it would be good to update the community on new
DTS developments since the last summit
* Ferruh asked on the mailing list about where DTS is for testsuites
which validate the flow api. Patrick responded stating that the 24.07
roadmap contains ethdev testsuites which we will work on soon. But, if
work moves fast and we can get time during the currently development
window, we can write some flow api testcases
* Honnappa notes that the end goal here is to demonstrate the
effectiveness of the framework and hand off the writing of testsuites
to maintainers/the community at large. So most of our efforts should
be towards that goal (don’t get distracted).
* Will email out the 24.07 roadmap once 24.03 releases
=====================================================================
Patch discussions
* Testpmd statefullness / params class
* EAL and testpmd params refactored to use this class
* Original idea was to set some defaults for testpmd to reduce some
load from developers. This does include these defaults.
* Mix between the existing defaults of testpmd and others in use
in DTS so far
* Luca is noting that default values on testpmd docs are
incorrect in some cases, some new params are not documented, etc.
* Do we expose ourselves to some risk by testpmd app changing,
and our testpmd params dataclass becomin “out of sync”
* May reduce developer user error by setting correct param names,
gives autocomplete to developers for params
* https://patchwork.dpdk.org/project/dpdk/list/?series=31622
* Scatter:
* Jeremy working on the secondary testcase which supports MLNX now
* Also working on the new testpmd context manager which will enable
having multiple testpmd testcases back to back in a testsuite (allows
for waiting for testpmd to fully close)
1. With most recent run on this patch, the new scatter suites is
running on mellanox NICs.
* Remove xmlrpc server implementation
* https://bugs.dpdk.org/show_bug.cgi?id=1374
* When receiving the packets to scapy on ingress port on tester we
need to encode the packets to some string so it can be consumed by the
dts runner. There is a method which translates bytes into base64
encoding, so now that we know about this method we should be all set.
* Can’t use a “scapy” interactive shell as we originally planned
due to issue with supporting cursor location, just using a python
shell which imports scapy *. Functionally this should be identical to
starting up a literal scapy shell.
* Improved Interactive shell output gathering
* https://patchwork.dpdk.org/project/dpdk/list/?series=31483
* Nick
* has tested his patch refactoring how users set hugepage from dts
or from default boot args
* He is also refactoring the config schema according to Juraj’s
comments, removing information which can be gathered from the system
1. One caveat is that Juraj suggested we don’t need to set os
value for nodes, but this is required to create the node’s os specific
remote session (which is needed to collect system info)
=====================================================================
Any other business
* UNH people are not changing ticket status when they start working on
a bugzilla ticket, which is confusing people. They will start with
this.
* Possible ticket: Check dependencies on remote nodes.
* Check python version on TG (required for scapy)
* Check requirements for DPDK, TG apps, etc. to ensure DTS can run
* Luca Vizzarrocan create a bugzilla ticket for this
