On Thu, Sep 04, 2025 at 06:57:52PM +0200, Pablo Neira Ayuso wrote:
> On Thu, Sep 04, 2025 at 06:21:59PM +0200, Phil Sutter wrote:
> > On Thu, Sep 04, 2025 at 05:24:59PM +0200, Pablo Neira Ayuso wrote:
> > > On Wed, Sep 03, 2025 at 07:22:59PM +0200, Phil Sutter wrote:
> > [...]
> > > > diff --git a/Makefile.am b/Makefile.am
> > > > index 5190a49ae69f1..9112faa2d5c04 100644
> > > > --- a/Makefile.am
> > > > +++ b/Makefile.am
> > > > @@ -23,6 +23,7 @@ libnftables_LIBVERSION = 2:0:1
> > > > ###############################################################################
> > > >
> > > > ACLOCAL_AMFLAGS = -I m4
> > > > +AM_DISTCHECK_CONFIGURE_FLAGS = --enable-distcheck
> > > >
> > > > EXTRA_DIST =
> > > > BUILT_SOURCES =
> > > > @@ -429,3 +430,11 @@ doc_DATA = files/nftables/main.nft
> > > > tools/nftables.service: tools/nftables.service.in ${top_builddir}/config.status
> > > > ${AM_V_GEN}${MKDIR_P} tools
> > > > ${AM_V_at}sed -e 's|@''sbindir''@|${sbindir}|g;s|@''pkgsysconfdir''@|${pkgsysconfdir}|g' <${srcdir}/tools/nftables.service.in >$@
> > > > +
> > > > +if !BUILD_DISTCHECK
> > > > +TESTS = tests/build/run-tests.sh \
> > > > + tests/json_echo/run-test.py \
> > > > + tests/monitor/run-tests.sh \
> > > > + tests/py/nft-test.py \
> > > > + tests/shell/run-tests.sh
> > > > +endif
> > > > diff --git a/configure.ac b/configure.ac
> > > > index da16a6e257c91..8073d4d8193e2 100644
> > > > --- a/configure.ac
> > > > +++ b/configure.ac
> > > > @@ -155,6 +155,11 @@ AC_CONFIG_COMMANDS([nftversion.h], [
> > > > AC_SUBST([MAKE_STAMP], ["\$(shell date +%s)"])
> > > > CFLAGS="${CFLAGS} -DMAKE_STAMP=\${MAKE_STAMP}"
> > > >
> > > > +AC_ARG_ENABLE([distcheck],
> > > > + AS_HELP_STRING([--enable-distcheck], [Build for distcheck]),
> > > > + [enable_distcheck=yes], [])
> > > > +AM_CONDITIONAL([BUILD_DISTCHECK], [test "x$enable_distcheck" = "xyes"])
> > >
> > > Oh no, with distcheck-hook: this is a lot cleaner.
> >
> > Hmm, I really don't see how it could be used for this purpose: It is
> > called before starting the VPATH build, here's an excerpt:
> >
> > | mkdir nftables-1.1.5/_build nftables-1.1.5/_build/sub nftables-1.1.5/_inst
> > | chmod a-w nftables-1.1.5
> > | test -d nftables-1.1.5/_build || exit 0; \
> > | dc_install_base=`CDPATH="${ZSH_VERSION+.}:" && cd nftables-1.1.5/_inst && pwd | sed -e 's,^[^:\\/]:[\\/],/,'` \
> > | && dc_destdir="${TMPDIR-/tmp}/am-dc-$$/" \
> > | && make distcheck-hook \
> > | && am__cwd=`pwd` \
> > | && CDPATH="${ZSH_VERSION+.}:" && cd nftables-1.1.5/_build/sub \
> > | && ../../configure \
> > | --enable-distcheck \
> > | \
> > | --srcdir=../.. --prefix="$dc_install_base" \
> > | && make \
> > | && make dvi \
> > | && make check \
> >
> > So by the time distcheck-hook runs, there is no Makefile(.in) I could
> > modify, only the top-level ones (which are tracked in git). What am I
> > missing here?
>
> distcheck-hook: could set a env var so test just print a [SKIP].
>
> Similar to your previous approach with the env var, but logic reversed.
I don't think the 'make distcheck-hook' call is able to inject variables
into the following 'make check' call's environment. It could create a
special file though which all test suites recognize and exit 77
immediately.
Cheers, Phil
