Bogdan <bogdro_...@gmx.us>, Fri Mar 03 2023 11:21:25 GMT+0100 (Central
European Standard Time)
Karl Berry <k...@freefriends.org>, Wed Mar 01 2023 20:01:26 GMT+0100
(Central European Standard Time)
Does anyone have access to an RHEL 8-based machine? Alma Linux,
Rocky Linux,
original RHEL, or even (sort of) CentOS 8? It would be nice if someone
could run a make check there (from automake dev).
git clone -q git://git.savannah.gnu.org/automake.git
cd automake
./bootstrap
./configure && make >&cm.out
make -j8 VERBOSE=1 check keep_testdirs=yes >&ch8.out
(choose whatever -j value you like)
[...]
This, along with my observation (below), MAY actually point to some
issue with Autoconf. Old systems work, new don't.
I take that back (for now). See below.
[...]
When I run
make check TESTS='t/nodef.sh t/backcompat2.sh t/backcompat3.sh
t/get-sysconf.sh t/lex-depend.sh t/remake-aclocal-version-mismatch.sh
t/subdir-add2-pr46.sh t/testsuite-summary-reference-log.sh'
I get failures in t/backcompat2.sh. Sometimes, like once-twice in
ten runs.
I've added a simple 'debug' in the form of the 'stat' command and:
--------------
[...]
--------------
Note that 'config.h' is older (4 seconds) than './configure', which
shouldn't be the case as it should get updated with new values.
What's funny, when I run just the failing test alone, it works 10
times in a row. A "Schroedinger bug" - when you try to observe, you
change the environment and the bug disappears :).
I'll try to do some more digging later.
Done a bit more digging into how AM_INIT_AUTOMAKE works. No idea how
or if caching of M4 'defines' works, but from a programming point of
view, I see setting the [no-define] option conditionally:
m4_ifval([$3], [_AM_SET_OPTION([no-define])])dnl
and checking it unconditionally:
_AM_IF_OPTION([no-define]
and I ask myself: what if on the next run the value is still
(un)defined when it shouldn't be? It shouldn't be defined, because we
start with the empty value as the third parameter in the
t/backcompat2.sh test, but...
Anyway, to the point: I did a simple change - un-define the value on
the "else" path. Diff attached (not pretty, because it's just for
testing).
Since the change, I've done 22 test runs of the same test set (the
one with 8 tests, not the single one which always passed) and saw
completely no failures.
If it works for everybody, then perhaps some _AM_UNSET_OPTION macro
should be created in options.m4, to keep the code nice, and that
committed to the main code instead of my patch.
If it works, there may be other places worth checking.
--
Regards - Bogdan ('bogdro') D. (GNU/Linux & FreeDOS)
X86 assembly (DOS, GNU/Linux): http://bogdro.evai.pl/index-en.php
Soft(EN): http://bogdro.evai.pl/soft http://bogdro.evai.pl/soft4asm
www.Xiph.org www.TorProject.org www.LibreOffice.org www.GnuPG.org
diff --git a/m4/init.m4 b/m4/init.m4
index 9b809ad5b..a79500695 100644
--- a/m4/init.m4
+++ b/m4/init.m4
@@ -62,7 +62,7 @@ dnl Distinguish between old-style and new-style calls.
m4_ifval([$2],
[AC_DIAGNOSE([obsolete],
[$0: two- and three-arguments forms are deprecated.])
-m4_ifval([$3], [_AM_SET_OPTION([no-define])])dnl
+m4_ifval([$3], [_AM_SET_OPTION([no-define])], [m4_ifdef(_AM_MANGLE_OPTION([no-define]), [m4_undefine(_AM_MANGLE_OPTION([no-define])])])dnl
AC_SUBST([PACKAGE], [$1])dnl
AC_SUBST([VERSION], [$2])],
[_AM_SET_OPTIONS([$1])dnl
