On Mon, 2023-09-04 at 11:06 +0200, Florian Westphal wrote:
> Running selftests on older kernels makes some of them fail very early
> because some tests use features that are not available on older
> kernels, e.g. -stable releases.
>
> Known examples:
> - inner header matching
> - anonymous chains
> - elem delete from packet path
>
> Also, some test cases might fail because a feature isn't
> compiled in, such as netdev chains for example.
>
> This adds a feature-probing to the shell tests.
On my Fedora 38, I have currently 7 tests failing.
With this patchset, tests/shell/testcases/maps/typeof_maps_add_delete
is now skipped (6 failing left).
Nice!
> +check_features()
> +{
> + for ffilename in $TESTDIR/../features/*.nft; do
> + feature=$(basename $ffilename)
> +
> + feature=${feature#*/}
> + feature=${feature%*.nft}
> +
> + eval NFT_HAVE_${feature}=0
> + $NFT --check -f "$ffilename" 2>/dev/null
> + if [ $? -eq 0 ]; then
> + eval NFT_HAVE_${feature}=1
the existing variables like VERBOSE,VALGRIND use "y" for true and
everything else is false.
I think 0|1 looks better. But it should be consistent. If 0|1 is used,
the other variables should be adjusted.
Note that on the other branch I added normalization functions
bool_y()/bool_n() to accept values like true/1/y/yes from the user, and
normalize to y|n. This could be changed internally to 1|0 without
breaking user setups.
> for testfile in $(find_tests)
> @@ -277,5 +296,10 @@ check_kmemleak_force
>
> msg_info "results: [OK] $ok [FAILED] $failed [TOTAL] $((ok+failed))"
>
> +if [ "$VERBOSE" == "y" ] ; then
> + echo "Probed Features:"
> + env | grep NFT_HAVE_
> +fi
xxx=$'\nNFT_HAVE_XXXXX=bogus' ./tests/shell/run-tests.sh /bin/true -v
gives the wrong output. Could be instead:
for v in $(compgen -v | grep '^NFT_HAVE_') ; do
echo "$v=${!v}";
done
