Closed MustafaFayez closed 2 years ago
Boost seems to be very sensitive to gcc version. Are you running the packages from a stock distro, or have you upgraded/downgraded?
I think that is the issue, I am running on a server, but I built boost yaml-cpp and libconfig from source to upgrade them. the stock distro versions were too old and showed compilation errors.
Edit: The OS is CentOS 7
Did you build boost and timeloop with the same compiler and library environment?
Yes, I used gcc/11.2 for both
Is there recommended versions of such libs to run on CentOS.
We are using boost 1.76.0 with gcc 11.2.0 in our internal CentOS 6 systems.
Honestly boost has been painful to deal with. The only reason we use it today is to dump out the stats/specs in XML format, which is then used by the regression scripts (via the Python parsers in the scripts/ directory).
If we had the time we would instead dump out the stats in YAML since yaml-cpp has been relatively painless, and rewrite the Python parser to read YAML instead of XML. Since this is an open-source project, we welcome contributions :-).
Agreed that boost is a pain :), I also have suffered with it recently. Migrating entirely to yaml-cpp seems like an attractive option.
I am currently talking to our server admins to make sure boost is packaged with a working version.
For anyone else facing something similar, I used a lightweight Arch Linux container named JuNest (Jailed User NEST) that allows to have disposable and partial isolated GNU/Linux environments within any generic GNU/Linux host OS and without the need to have root privileges for installing packages. https://github.com/fsquillace/junest I used it because some system admins don't allow Docker due to security issues. It works well for me, and is pretty easy to use.
I see this error while running timeloop:
the libs versions are boost 1.8.0 libconfig 1.7.3 gcc 11.2