> I like to check in early and often, but my TaskManager performance project > has an experimental aspect to it. The TaskManager changes should increase > concurrency, which may shake out bugs in other code, as well as having their > own risks. Only benchmarking can prove that a change improves performance - > intuition and experience are notoriously unreliable predictors. >
Could not agree more. > The TaskManager changes should be tested against all runnable tests before > being checked into the trunk, but that can only be done if they are based on > a version that passes all runnable tests. > See my suggestions in previous email. We should first and foremost establish a solid baseline that passes as much tests as we can throw at it. > That all seems to me to be a job for branches. > We have "skunk" branches for experimentation.... In fact there is a skunk branch right now called "NewTaskManager". This branch should be based off a stable release indeed (see above), not sure if that is the case with the skunk/NewTaskManager branch. > Patricia >
