Thanks for finding this, I agree that it seems to fail due to the
number of cores as you suggested:

...
obj-x86_64-linux-gnu/app/test/dpdk-test '-l 0-143'
...
| EAL: Detected 128 lcore(s)
...
| EAL: invalid core list syntax


But when trying to recreate this I've found the -l argument to not
even be passed to the tests.
Neither with the coming 22.11 which we work on, nor with 21.11-5
(unstable) nor with 22.11.6 (stable).

The code in app/test/meson.build does not really add anything in
between the binary and the no-huge argument:
 463 dpdk_test = executable('dpdk-test',
 483     test_args = []
 487             test_args += test_no_huge_args

Also in all other builds there is no -l at all:
"... app/test/dpdk-test --no-huge -m 2048 --file-prefix=bitmap_autotest"
https://buildd.debian.org/status/fetch.php?pkg=dpdk&arch=amd64&ver=21.11-5%2Bb2&stamp=1667749698&raw=0
https://launchpadlibrarian.net/623152077/buildlog_ubuntu-kinetic-amd64.dpdk_21.11.2-0ubuntu1_BUILDING.txt.gz


Therefore I wanted to ask if you could execute the very same you did
on a smaller host and let us know if then the -l in the test calls is
gone or if you see it still there but with smaller values matching
your build environment?

If yes, then we need to assume that the environment size somehow
causes this arg to be added, but otherwise it stays a mystery.

--
Christian Ehrhardt
Senior Staff Engineer, Ubuntu Server
Canonical Ltd

Reply via email to