Hi everyone,
We hit a bug in stunnel-5.05 and it continues in 5.06. The problem comes from systemd which moved sd_listen_fds() form libsystemd-daemon.so to libsystemd.so. However, stunnel's configure.ac still has
AC_SEARCH_LIBS([sd_listen_fds], [systemd-daemon])
For backwards compatibility, this should probably be
AC_SEARCH_LIBS([sd_listen_fds], [systemd], [], [ AC_SEARCH_LIBS([sd_listen_fds], [systemd-daemon]) ])
or similar. Please refer to our downstream bug:
https://bugs.gentoo.org/show_bug.cgi?id=525234'
Thanks.
-----BEGIN PGP SIGNED MESSAGE----- Hash: SHA1
Anthony G. Basile wrote:
We hit a bug in stunnel-5.05 and it continues in 5.06. The problem comes from systemd which moved sd_listen_fds() form libsystemd-daemon.so to libsystemd.so.
Not really a bug, but rather missing support for systemd 209 or later: http://lists.freedesktop.org/archives/systemd-devel/2014-February/017146.htm...
We merged libsystemd-journal.so, libsystemd-id128.so, libsystemd-login and libsystemd-daemon into a a single libsystemd.so to reduce code duplication and avoid cyclic dependencies (see below).
Please try: https://www.stunnel.org/downloads/beta/stunnel-5.07b2.tar.gz
Mike
On 10/16/14 00:46, Michal Trojnara wrote:
-----BEGIN PGP SIGNED MESSAGE----- Hash: SHA1
Anthony G. Basile wrote:
We hit a bug in stunnel-5.05 and it continues in 5.06. The problem comes from systemd which moved sd_listen_fds() form libsystemd-daemon.so to libsystemd.so.
Not really a bug, but rather missing support for systemd 209 or later: http://lists.freedesktop.org/archives/systemd-devel/2014-February/017146.htm...
We merged libsystemd-journal.so, libsystemd-id128.so, libsystemd-login and libsystemd-daemon into a a single libsystemd.so to reduce code duplication and avoid cyclic dependencies (see below).
Please try: https://www.stunnel.org/downloads/beta/stunnel-5.07b2.tar.gz
Mike -----BEGIN PGP SIGNATURE----- Version: GnuPG v1
iEYEARECAAYFAlQ/TcMACgkQ/NU+nXTHMtE0oACeNbCPKzPZ7T5RiL6d8Cyff/u0 hqsAn1w2GEGxEv39NZEUTf6ZKltRx9Ls =AM7Z -----END PGP SIGNATURE----- _______________________________________________ stunnel-users mailing list [email protected] https://www.stunnel.org/cgi-bin/mailman/listinfo/stunnel-users
Hi Michal,
That patch works, but we prefer in Gentoo to have a configuration switch to avoid automatic dependencies [1]. I've attached a patch against configure.ac that would --enable/disable-systemd socket activation. It applies against 5.07b2.
Ref. [1] https://wiki.gentoo.org/wiki/Project:Quality_Assurance/Automagic_dependencie...
-----BEGIN PGP SIGNED MESSAGE----- Hash: SHA1
Anthony G. Basile wrote:
That patch works, but we prefer in Gentoo to have a configuration switch to avoid automatic dependencies [1]. I've attached a patch against configure.ac that would --enable/disable-systemd socket activation. It applies against 5.07b2.
Hi Anthony,
You have made a very good point. Automagic dependencies are evil indeed. I have added "--disable-systemd" option to the ./configure script.
Please try: https://www.stunnel.org/downloads/beta/stunnel-5.07b3.tar.gz
Mike
On 10/23/14 11:24, Michal Trojnara wrote:
-----BEGIN PGP SIGNED MESSAGE----- Hash: SHA1
Anthony G. Basile wrote:
That patch works, but we prefer in Gentoo to have a configuration switch to avoid automatic dependencies [1]. I've attached a patch against configure.ac that would --enable/disable-systemd socket activation. It applies against 5.07b2.
Hi Anthony,
You have made a very good point. Automagic dependencies are evil indeed. I have added "--disable-systemd" option to the ./configure script.
Please try: https://www.stunnel.org/downloads/beta/stunnel-5.07b3.tar.gz
Mike -----BEGIN PGP SIGNATURE----- Version: GnuPG v1
iEYEARECAAYFAlRJHb4ACgkQ/NU+nXTHMtGgtQCdG4my8GP+eireI6yyC1zS1+CG 0yEAoKC6bp+Ni/s7SWsJjjrHL258aoBN =4Nij -----END PGP SIGNATURE----- _______________________________________________ stunnel-users mailing list [email protected] https://www.stunnel.org/cgi-bin/mailman/listinfo/stunnel-users
Hi Mike,
On a system without systemd installed `./configure --enable-systemd` will set the following in src/config.h
#undef HAVE_SYSTEMD_SD_DAEMON_H *
#define USE_SYSTEMD 1
leading to a compile-time failure. My original suggestion didn't have that. It comes from the unconditional AC_DEFINE([USE_SYSTEMD], [1] ... at line 391 of configure.ac. If you don't like the `if test $ac_cv_search_sd_listen_fds = "no"; then` you can repeat your logic from line 404 at 391 to fix this:
AC_SEARCH_LIBS([sd_listen_fds], [systemd systemd-daemon], AC_CHECK_HEADERS([systemd/sd-daemon.h], [ AC_DEFINE([USE_SYSTEMD], [1],
-----BEGIN PGP SIGNED MESSAGE----- Hash: SHA1
Anthony G. Basile wrote:
On a system without systemd installed `./configure --enable-systemd` will set the following in src/config.h
#undef HAVE_SYSTEMD_SD_DAEMON_H *
#define USE_SYSTEMD 1
leading to a compile-time failure.
This is precisely the intended behavior. If a user forcefully enabled a feature on a system that does not support this feature, stunnel should *not* silently ignore the user's choice and build cleanly. http://en.wikipedia.org/wiki/Principle_of_least_astonishment
Otherwise, the user might spend hours trying to find out why the feature she they explicitly enabled does not work.
Mike
-
Anthony G. Basile -
Michal Trojnara