Description of problem: On a fresh install of F16, if I install aeolus-all (which includes iwhd), and then run aeolus-configure, iwhd fails to start: [root@localhost ~]# sudo service iwhd status iwhd.service - Image Warehouse Loaded: loaded (/lib/systemd/system/iwhd.service; enabled) Active: failed since Wed, 22 Feb 2012 14:58:47 -0500; 12min ago Process: 3652 ExecStart=/usr/bin/iwhd $IWHD_ARGS (code=dumped, signal=SEGV) CGroup: name=systemd:/system/iwhd.service There is no /var/log/iwhd.log file. Version-Release number of selected component (if applicable): aeolus-all.noarch 0.4.0-1.fc16 iwhd.x86_64 1.2-1.fc16 How reproducible: 100% for me and also jclift Steps to Reproduce: 1. Provision F16 installation 2. yum install aeolus-all 3. aeolus-configure # will fail with errors starting iwhd 4. service iwhd status Actual results: iwhd fails with an error hinting at some sort of segfault Expected results: iwhd starts normally Additional info: I haven't seen this happen on other platforms, so this has probably been fixed long ago and just not included in the Fedora 16 packages. This is a big problem, though, because it keeps Aeolus from being usable to new users on F16.
To my knowledge F16 is not on the list of supported platforms for CF. So this shouldn't be treated as a blocker. That said, there's no reason for it to be gratuitously broken. Pete?
Thanks for the report, Matt. This appears to be due to changes in boost. iwhd doesn't even get through its configure script on F16, now. It dies because the latest boost headers now choke the C++ preprocessor with an #error if you're not compiling with -pthread. Our test for whether to use -pthread involves compiling without the option for which we're probing. This appears related: https://svn.boost.org/trac/boost/ticket/6165 I'll dig more tomorrow.
The build failures I mentioned arise only when using g++-4.7.x. With F16's /usr/bin/g++, iwhd does still build. However, the result fails each and every mongo-using test. The problem there is that mongod was taking more than the 5-seconds the test allotted for it to start up. In practice, with the new DB initialization it does, it now takes ~15 seconds to become usable, and far more when you run the tests in parallel and it's trying to initialize 5 or 10 separate DBs at once. In addition, the disk space requirements have now gone way up, especially when running the tests in parallel. For example, "make -j1 check" passes once I've increased the timeout value from 5 to 30 seconds, but several tests fail in mysterious ways when you do "make -j4 check".
Hi Jim, For what it's worth (and if you're not already aware of this), this appears to work just fine with iwhd-1.2-3 available in the Aeolus testing repo: http://repos.fedorapeople.org/repos/aeolus/conductor/testing/fedora-16/x86_64/ It's just the version available in native F16 that does not work.
Hi guys, Can we get iwhd-1.2-3 into mainline Fedora 16 updates? It does work, unlike the version presently there. :) If not, what's the way forward? End users coming from Fedora 16 are completely unable to use Aeolus (ie blocker) due to this. :(
I'm about to release 1.4, which includes a fix for 798786, so expect it very soon, probably today.
this is now fixed in F16
[root@nec-em18 ~]# service iwhd status iwhd.service - Image Warehouse Loaded: loaded (/lib/systemd/system/iwhd.service; enabled) Active: active (running) since Tue, 27 Mar 2012 11:08:45 -0400; 17min ago Main PID: 2753 (iwhd) CGroup: name=systemd:/system/iwhd.service └ 2753 /usr/bin/iwhd -c /etc/iwhd/conf.js -l /var/log/iwhd.log [root@nec-em18 ~]# rpm -q aeolus-all aeolus-all-0.9.0-0.20120326215437gitd951bad.fc16.noarch [root@nec-em18 ~]# rpm -q iwhd iwhd-1.4-1.fc16.x86_64 [root@nec-em18 ~]# cat /etc/redhat-release Fedora release 16 (Verne) successfully passed aeolus-configure. Moving bug to VERIFIED
This is a fedora bug, removing from the product list.