On Fri, Feb 25, 2022 at 03:39:02PM +0100, Aleksandar Lazic wrote: > > > diff --git a/reg-tests/sample_fetches/srv_rtt.vtc > > > b/reg-tests/sample_fetches/srv_rtt.vtc > > > new file mode 100644 > > > index 000000000..c0ad0cbae > > > --- /dev/null > > > +++ b/reg-tests/sample_fetches/srv_rtt.vtc > > > @@ -0,0 +1,34 @@ > > > +varnishtest "srv_rtt sample fetch Test" > > > + > > > +#REQUIRE_VERSION=2.6 > > > > Note, we *might* need to add a new macro to detect support for TCP_INFO. > > We still don't have the config predicates to detect support for certain > > keywords or sample fetch functions so that's not easy, but it's possible > > that this test will break on some OS like cygwin. If so we could work > > around this temporarily using "EXCLUDE_TARGETS" and in the worst case we > > could mark it broken for the time it takes to completely solve this. > > Agree here the suggestion is to add something like USE_GETSOCKOPT to be able > to make '#REQUIRE_OPTIONS=GETSOCKOPT', something like similar to > USE_GETADDRINFO, right?
No, it's not even this, it would in fact need to detect TCP_INFO which comes from the system's includes. That's why we won't do that and instead we'll later finish to add more predicates to detect supported config keywords. This way if we mention a dependency on "bc_rtt", we know we can skip this test on systems which don't include it based on haproxy's internal features detection at build time. Do not worry for this one, leave it as-is, we'll see what breaks on the CI (likely only cygwin) and we'll exclude it/them. Cheers, Willy
