Red Hat Bugzilla – Bug 994725
Compiling autoconf-2.69-13.fc20.src.rpm fails one of the self tests
Last modified: 2014-01-10 10:08:05 EST
Created attachment 784138 [details] Test log that contains output of internal testing while building rpm Description of problem: I'm compiling the autoconf-2.69-13.fc20.src.rpm. Everything works except one check. The output says to send the failure to GNU, but I thought I would give it to you since it is a Fedora package. Autoconf base layer. 216: AC_REQUIRE: topological sort ok 217: AC_REQUIRE: error message ok 218: AC_REQUIRE & AC_DEFUN_ONCE: Require, expand ok 219: AC_REQUIRE & AC_DEFUN_ONCE: Expand, require ok 220: AC_REQUIRE & AC_PROVIDE ok 221: AC_INIT ok 222: AC_INIT with unusual version strings ok 223: AC_COPYRIGHT ok 224: AC_CACHE_CHECK skipped (base.at:488) 225: AC_CACHE_LOAD ok 226: AC_COMPUTE_INT ok 227: AC_TRY_COMMAND ok 228: Input/Output ok 229: configure arguments ok 230: AC_ARG_ENABLE and AC_ARG_WITH ok 231: configure directories FAILED (base.at:705) Please send `tests/testsuite.log' and all information you think might help: To: <bug-autoconf@gnu.org> Subject: [GNU Autoconf 2.69] testsuite: 231 failed Version-Release number of selected component (if applicable): autoconf-2.69-13.fc20.src.rpm How reproducible: I only tried it once. I assume it would fail every time. Steps to Reproduce: 1.Build the rpm from the src.rpm package, rpmbuild -bb autoconf.spec 2. 3. Actual results: Fails test 231 Expected results: Passes all tests Additional info:
Thanks for the report. I'll raise this upstream to fix the testsuite - but it is not worth to fix downstream. The problem is: a) you have fedora 17 b) you have installed autoconf from rawhide (including config.site enabled) c) you are building it with old redhat-rpm-macros (which does not set the CONFIG_SITE=NONE before building the package). Anyway, upstream testsuite should take the config.site file into account. I'll close this once fixed in upstream. Pavel
Thanks for your explanation. I knew I was stepping outside the lines, but this confirms it. Reports of confusing anaconda functionality for my use case scared me off F19. So I'm waiting for F20 alpha, with the accompanying anaconda improvements, to install F20 and thought I would tweak F17 in the meantime. I've run into other areas where incompatibility has been more obvious. Still, a learning experience. I have no internet facing services, so this isn't as reckless as it sounds. :-)
This bug appears to have been reported against 'rawhide' during the Fedora 20 development cycle. Changing version to '20'. More information and reason for this action is here: https://fedoraproject.org/wiki/BugZappers/HouseKeeping/Fedora20
Upstream report: http://lists.gnu.org/archive/html/autoconf-patches/2013-11/msg00002.html
I'd like to request one slight improvement - to not redefine CONFIG_SITE in case it is already defined. Consider that you wan't to let your compiler know if it is run from ./configure script during rpmbuild (compiler may be run from multiple places in spec otherwise, especially by make command). The config.site would be very useful place to export some variable for that. diff --git a/macros.in b/macros.in index e073048..b0c6d6f 100644 --- a/macros.in +++ b/macros.in @@ -654,7 +654,7 @@ package or when debugging this package.\ export CLASSPATH}\ PKG_CONFIG_PATH=\"${PKG_CONFIG_PATH}:%{_libdir}/pkgconfig:%{_datadir}/pkgconfig\"\ export PKG_CONFIG_PATH\ - CONFIG_SITE=NONE\ + : ${CONFIG_SITE:=NONE}\ export CONFIG_SITE\ \ %{verbose:set -x}%{!verbose:exec > /dev/null}\
Scratch the previous comment, it is not about this bugzilla at all, sorry.
Fixed upstream similarly like in the comment #4: http://pkgs.fedoraproject.org/cgit/autoconf.git/commit/?id=50451905
Damn..., again bad link, sorry. Here is the right one: http://git.savannah.gnu.org/cgit/autoconf.git/commit/?id=251d13ff24