Rerun Modes

There are several "modes" of running (retracing), having to do with the way jobs are scheduled and the way jobs are executed. The default mode of execution is "SAFE".

This flag affects the execution of the jobs. If the execution is FAST, runtime tracing is disabled for the job, meaning that the dependencies are not recomputed.
This flag affects the firing of the jobs in the presence of barriers. Normally, a job is ready to fire only if its entire up-cone is VALID. If the "AGGRESSIVE" flag is used, a job will fire if all its immediate inputs are VALID. The situation where the immediate inputs are VALID but the up-cone is not VALID normally arises when you have barriers.

In AGGRESSIVE mode, jobs are fired in parallel above and below the barrier. If the barrier stops the change propagation, all is OK. Otherwise, the jobs below the barrier will have to be re-executed.

The default mode is not AGGRESSIVE, meaning that all jobs above the barrier are executed before the jobs below the barrier.

This flag affects the running of jobs in the presence of barriers. Normally, when you run a target, only the jobs below the barriers are run. With the "CROSSBARRIERS" flag, all jobs in the up-cone of the target will be run.
This flag affects the running of FAILED jobs. Normally, FAILED jobs remain FAILED, until the inputs change. With the "FORCEFAILED" flag, the failed nodes in the up-cone of the target are also run.
This flag affects the checking of the primary inputs in the up-cone of the target. Normally, FlowTracer checks the timestamp of all primary inputs, but this could be expensive especially if you are using a slow filesystem such as ClearCase. If you set this flag, the check is skipped.

This flag reduces the reliability of the running. You should use this flag in conjunction with alternative ways to check timestamps, such as vls or vovcheckfiles.

This is the name for the default mode, the one in which all flags mentioned above are OFF.