First, thanks for the lightning fast response to my first petsc-dev question (or is it negative time, since Matt had already implemented what I asked for last night).
My next configure wish list thing is to have a way to "restart" configure from a certain point. This is probably a bit tricky to implement. The syntax I am thinking of (and there may be better options) would be to have a --resume or --continue option given to configure, in addition to new or modified options. Then, it would restart from the "right" place, given the new set of options, while avoiding redundant checks. For example, if there was a bad mpi path given to petsc's configure, I would like to be able to rerun it with a corrected mpi location (everything else staying the same) and avoid rerunning all the other unrelated configuration checks that happen to come before it. Obviously this saves a lot of time for failures that occur near the end of configure. This is not like gnu configure's caching, which doesn't really work anyway. I'm not sure there is enough infrastructure right now to implement something like this; for example, the dependencies between various configure options are not apparent. I don't think, however, that a completely general solution is essential -- even something as simple as restarting at the last error would be great. In any case I think this would be a "use at your own risk" option since there will always be the fallback full rerun of configure. Boyana
