Does the evidence in favour of code reviews as a quality technique apply to reviewing pull requests? It seems to me that looking at code through the letterbox view of diffs isn't effective at spotting non-local problems.
-
I haven't got a huge amount of experience of them, but I think that if the reviewer doesn't know the original code well enough to begin with (as may often be the case), the utility of the review is limited to the superficial and local.
I've often found myself doing reviews where I'm raising petty (but valid) issues like "you could choose a better identifier name", and then qualifying the entire review by saying that I've no idea if it actually stands a chance of working.