Jeff King <p...@peff.net> writes: > Yeah. We tolerate a certain amount of skew (24 hours for --name-rev, and > 5 broken commits in a row for --since). But the big ones are usually > software bugs (the big kernel ones were from broken "guilt", I think) or > broken imports (when I published a bunch of skew statistics last year, > the interesting ones were all imports; I don't know if they were > software bugs, or just garbage in, garbage out).
I was hoping that 2e6bdd3 (test-generation: compute generation numbers and clock skews, 2012-09-04) may be a good first step to come up with a practical and cheap solution on top of it. The traversal can be fooled by clock skews when it sees a commit that has a timestamp that is older than it should, causing it to give up, incorrectly thinking that there won't be newer commits that it is interested in behind the problematic commit. The logic implemented by the change is to identify these problematic commits, and we could record these commits with the value of the timestamps they should have had (e.g. the timestamp of the newest ancestor for each of these commits) in a notes tree. Then the traversal logic (commit-list-insert-by-date) could be updated use that "corrected" timestamp instead not to be fooled by the clock skew. Such a notes tree can be built once and updated by only "appending", as a commit will never acquire more ancestors in its parents chain once it is made. Is it too simplistic, or too costly? In git.git we have three such commits whose timestamp need to be corrected, while in the Linux kernel there were 2.2k skewed commits when I counted them a few months ago. -- To unsubscribe from this list: send the line "unsubscribe git" in the body of a message to majord...@vger.kernel.org More majordomo info at http://vger.kernel.org/majordomo-info.html