On 2/17/21 11:25 AM, Thomas Huth wrote:
> Even though the host machines that run the Travis CI jobs have
> quite a lot of CPUs (e.g. nproc in an aarch64 job reports 32), the
> containers on Travis are still limited to 2 vCPUs according to:
>
> https://docs.travis-ci.com/user/reference/overview/#approx-boot-time
>
> So we do not gain much when compiling with a job number based on
> the output of "getconf _NPROCESSORS_ONLN" - quite the contrary, the
> aarch64 containers are currently aborting quite often since they
> are running out of memory. Thus let's rather use a fixed number
> like 3 in the jobs here, so that e.g. two threads can actively run
> while a third one might be waiting for I/O operations to complete.
> This should hopefully fix the out-of-memory failures in the aarch64
> CI jobs.
>
> Signed-off-by: Thomas Huth <th...@redhat.com>
> ---
> See also this URL for the explanation of the 32 vs 2 CPUs on aarch64:
> https://travis-ci.community/t/nproc-reports-32-cores-on-arm64/5851
>
> .travis.yml | 2 +-
> 1 file changed, 1 insertion(+), 1 deletion(-)
>
> diff --git a/.travis.yml b/.travis.yml
> index fc27fd6330..cc39a447e8 100644
> --- a/.travis.yml
> +++ b/.travis.yml
> @@ -88,7 +88,7 @@ git:
> # Common first phase for all steps
> before_install:
> - if command -v ccache ; then ccache --zero-stats ; fi
> - - export JOBS=$(($(getconf _NPROCESSORS_ONLN) + 1))
> + - export JOBS=3
Maybe worth a comment around?
Reviewed-by: Philippe Mathieu-Daudé <phi...@redhat.com>
> - echo "=== Using ${JOBS} simultaneous jobs ==="
>
> # Configure step - may be overridden
>
