51 ms·
Don't disagree on much. What if the cost of backing out a change actually was zero or close to zero? What if the prospective change actually was run on all pla
by sitkack 11y ago
Don't disagree on much.
What if the cost of backing out a change actually was zero or close to zero? What if the prospective change actually was run on all platforms with all of the historic input the function has ever seen?
What happens when our undo buffer and our log of refactorings gets checked into version control. Not a log of pure text, but a log of semantic transformations.
Of course bad patches make it into the tree. Right now all the bad patches require human intervention. I am saying that it might be economically advantageous to let in some of the bad patches automatically so it takes less work.
For most codebases I see in industry, simply rejecting a patch on lint failure, lack of tests or documentation would effectively freeze most trees.