[ajross]

The kernel folks shoot people regularly for far milder offenses, so that's not really surprising. That said, I don't think we're reading commit logs for the same kernel. I see two-liners routinely in the log (because checkpatch wants the shortlog and a body -- but note that these are often just rephrasings of each other). And the kernel workflow encourages batching up what would otherwise be many small changes into well-formed patch sets, which means that you get a coarser granulatiry of changes than many projects.

And sure, there are patches that go in with elaborate analysis in the commit log. Where those provide good information, I think they're great. Most of the time, they just amount to a BUG() trace stuffed in for no good reason and I find them to be noise.

And I don't see how git blame helps my problem. I have a regression, and an intuition of what kinds of changes might have caused it. A commit log will tell me whether a commit seems like a good candidate to inspect. Git blame just tells me who touched specific lines. If I knew the specific lines to look at, I wouldn't need to be applying intuition here, I'd just fix the bug.