no-re...@cfengine.com writes:
> I am having the same issues with this version running on SLES 11 x86_64 (phys
> 2.6.27.29-0.1-default). It only occurs every day or so.
> Frans, did you ever find out what was happening?
>
>
> It is statically compiled:
>
> BerkeleyDB
> tar xvfz db-5.0.26.tar.gz && cd db-5.0.26/build_unix
> ../dist/configure --prefix=/tmp/cf_d --enable-static --disable-shared
> make && make install
> cd ../..
>
> OpenSSL
> tar xvfz openssl-1.0.0a.tar.gz && cd openssl-1.0.0a
> ./config --prefix=/tmp/cf_d no-shared no-dso
> make && make install
> cd ../..
>
> PCRE
> tar xvfz pcre-8.10.tar.gz && cd pcre-8.10
> ./configure --enable-static --disable-shared --prefix=/tmp/cf_d
> --enable-unicode-properties
> make && make install
> cd ../..
>
> Cfengine 3
> tar xvfz ccfengine-3.0.5p1.tar.gz && cd cfengine-3.0.5p1.tar.gz
> ./configure --with-openssl=/tmp/cf_d --with-pcre=/tmp/cf_d
> --with-berkeleydb=/tmp/cf_d --enable-static --disable-shared
> LDFLAGS='--static'
> make && make install
We had exactly the same problem. On some architectures, it segfaulted
more or less occasionally, while on others (rhel6 beta) the segfaults
were consistent. Our solution was to drop static compilation of the
cfengine binaries. The libs are still linked statically, so the only
libs the binaries depends on are the glibc libs:
[r...@rhel6-test1 ~]# ldd /var/cfengine/bin/cf-agent
linux-vdso.so.1 => (0x00007fff40916000)
libnss_nis.so.2 => /lib64/libnss_nis.so.2 (0x00007f01155b7000)
libpthread.so.0 => /lib64/libpthread.so.0 (0x000000365e000000)
librt.so.1 => /lib64/librt.so.1 (0x000000365e800000)
libm.so.6 => /lib64/libm.so.6 (0x000000365ec00000)
libc.so.6 => /lib64/libc.so.6 (0x000000365d800000)
libnsl.so.1 => /lib64/libnsl.so.1 (0x0000003661000000)
libnss_files.so.2 => /lib64/libnss_files.so.2 (0x00007f01153a8000)
/lib64/ld-linux-x86-64.so.2 (0x000000365d400000)
The idea is that if glibc is broken, everything is broken and the
solution is a complete reinstall anyway. If you drop the
"LDFLAGS='--static'" from the cfengine compilation the binaries should
stop segfaulting. The binaries will link dynamically against glibc but
that is acceptable IMHO.
Cheers,
--
Trond H. Amundsen <t.h.amund...@usit.uio.no>
Center for Information Technology Services, University of Oslo
_______________________________________________
Help-cfengine mailing list
Help-cfengine@cfengine.org
https://cfengine.org/mailman/listinfo/help-cfengine
