[dpdk-dev] Non-deterministic build failure in rte_ethdev.h:4287:10 on Taishan ARM server

["Juraj Linkeš" <juraj.linkes@pantheon.tech>]

Hi DPDK devs,

I'm hitting a build failure with dpdk 20.02.1 with rte_ethdev.h that's happening seemingly at random:
make install T=arm64-armv8a-linuxapp-gcc -j
In file included from /tmp/openvpp-testing/dpdk/lib/librte_ethdev/rte_class_eth.c:13:0:
/tmp/openvpp-testing/dpdk/lib/librte_ethdev/rte_ethdev.h:4287:10: fatal error: rte_ethdev_core.h: No such file or directory
#include <rte_ethdev_core.h>
          ^~~~~~~~~~~~~~~~~~~
compilation terminated.
make[5]: *** [rte_class_eth.o] Error 1
make[5]: *** Waiting for unfinished jobs....
make[4]: *** [librte_ethdev] Error 2
make[3]: *** [lib] Error 2
make[2]: *** [all] Error 2
make[1]: *** [pre_install] Error 2
make: *** [install] Error 2

Full log here: https://logs.fd.io/production/vex-yul-rot-jenkins-1/csit-dpdk-perf-verify-master-3n-tsh/11/archives/log.html.gz

This doesn't happen every time. It took me five builds on two machines (for a total of 10 builds) to hit this issue, but it did eventually happen. I tried this with 20.02 at first and it seemed to be happening more frequently, but I don't have enough data to really tell if there's an actual difference in frequency.

We're not doing anything special, just downloading and extracting the archive, then setting CONFIG_RTE_LIBRTE_MLX5_PMD and CONFIG_RTE_LIBRTE_I40E_16BYTE_RX_DESC to y in config/common_base and then running make install T=arm64-armv8a-linuxapp-gcc -j. As mentioned in the subject, the build server is a Taishan ARM server.

Any ideas on how to fix (or workaround) this? Is there a chance meson build would fix this? If so, how do we make sure we build testpmd with mxl5 support and how do we set CONFIG_RTE_LIBRTE_I40E_16BYTE_RX_DESC with meson build?

Thanks,
Juraj