* Stefano Lattarini wrote on Mon, Mar 14, 2011 at 12:51:29PM CET:
> > Or maybe we could define a "maintainer-check" that looks at the
> > testcases' names and complains about suspicious ones... I'll
> > give it a shot unless someone beats me.
> >
> Done in the attached patch.
OK with nits addressed.
Thanks,
Ralf
> Subject: [PATCH] maintcheck: look for problematic names of testcases
>
> The configure.in stub created by default by `tests/defs' obtains
> the first argument of AC_INIT from the test name, and this can
> cause some supported autoconf versions to fail with a spurious
> error if that test name contains the name of an m4 builtin (e.g.,
> `defn' or `undefine').
I'd say "name of a macro", because I don't think it's only builtins that
are affected. For example, AC_DEFUN wouldn't work either. So, it would
be good to also test that we don't use prefixes _?A[CUMHS]_ or _?m4_
either (can be done in followup patch).
> See for example the bug fixed by commit v1.11-287-g1325a8a.
>
> This change add a maintainer check that warns about test names
> which are possibly problematic in this regard.
>
> * Makefile.am (sc_test_names): New maintainer-check target.
> (syntax_check_rules): Add it.
> (m4_builtins): New helper variable.
> (TESTS): Updated according to the following renamings.
> * tests/include.test: Renamed ...
> * tests/hdr-vars-defined-once.test: ... to this.
> * tests/sinclude.test: Renamed ...
> * tests/m4-inclusion.test: ... to this, and simplified
> accordingly.
> * tests/include2.test: Renamed ...
> * tests/dist-included-parent-dir.test: ... to this, for
> consistency.
> --- a/Makefile.am
> +++ b/Makefile.am
> @@ -111,6 +111,7 @@ dist-hook:
> # Some simple checks, and then ordinary check. These are only really
> # guaranteed to work on my machine.
> syntax_check_rules = \
> +sc_test_names \
> sc_diff_automake_in_automake \
> sc_perl_syntax \
> sc_no_brace_variable_expansions \
> @@ -151,6 +152,61 @@ sc_at_in_texi
> $(syntax_check_rules): automake aclocal
> maintainer-check: $(syntax_check_rules)
>
> +## Look for test whose names can cause spurious failures when used as
> +## first argument to AC_INIT (chiefly because they might contain an
> +## m4/m4sugar builtins).
either "builtin" or s/ an//
> +m4_builtins = \
> + bpatsubst \
> + bregexp \
> + builtin \
> + changecom \
> + changequote \
> + changeword \
> + debugfile \
> + debugmode \
> + decr \
> + define \
> + defn \
> + divert \
> + divnum \
> + dumpdef \
> + errprint \
> + esyscmd \
> + eval \
> + format \
> + ifdef \
> + ifelse \
> + include \
> + incr \
> + index \
> + indir \
> + len \
> + maketemp \
> + mkstemp \
> + patsubst \
> + popdef \
> + pushdef \
> + regexp \
> + shift \
> + sinclude \
> + substr \
> + symbols \
> + syscmd \
> + sysval \
> + traceoff \
> + traceon \
> + translit \
> + undefine \
> + undivert
> +sc_test_names:
> + @m4_builtins_rx=`echo $(m4_builtins) | sed 's/ /|/g'`; \
> + m4_builtins_rx="\\<($$m4_builtins_rx)\\>|\\<_?m4[a-zA-Z0-9_]"; \
^
This underscore plus question doesn't make sense to me. Why not this?
\\<m4_[a-zA-Z0-9]
> + if ls tests/*.test | LC_ALL=C grep -E "$$m4_builtins_rx"; then \
> + echo "the names of the tests above can be problematic" 1>&2; \
> + echo "Avoid test names that contain names of m4 builtins" 1>&2; \
> + exit 1; \
> + fi
> +
> ## This check avoids accidental configure substitutions in the source.
> ## There are exactly 8 lines that should be modified. This works out
> ## to 28 lines of diffs.
> diff --git a/tests/include2.test b/tests/dist-included-parent-dir.test
> diff --git a/tests/include.test b/tests/hdr-vars-defined-once.test
> diff --git a/tests/sinclude.test b/tests/m4-inclusion.test
