Summary
- prefer old-style import "=>";
- merged
- redundant (see also 3069da1743bc);
- removed obsolete table (see also 6acd1a2bd146);
- more robust init_database(); proper treatment of views, not tables (amending dd350a41594c);
- proper private_data.transaction_lock; prefer execute_batch_statement;
- clarified names;
- proper support for SSH;
- tuned signature;
- tuned imports;
- add module for faster scheduled builds;
- always use host database and make protected;
- read relative cpu from build log;
- prefer extensible next_node_info in build process over process_options in build engine (which needs the final node info anyway);
- added start date to build jobs, e.g., for build time estimation;
- added initial version of benchmark module, e.g., to compare performance of different hosts; added benchmark operation to build cluster;
- generalized node infos: allow addressing of numa node segments via relative cpus; add more node options and process policy options using taskset as alternative to NUMA for more fine-grained cpu controls (e.g., for cpus with heterogeneous cores in the same NUMA segment);
- add information about static host resources such as available NUMA nodes, cpus, etc. (e.g., for offline build planning);
- added Range object to Host, e.g. to parse/unparse NUMA node ranges;
- defined statically known tables of Build_Log; read hostname from build logs, store in Session_Entry (e.g., to track hosts in distributed build);