On Thu, Oct 17, 2019 at 05:06:54PM -0700, Linus Torvalds wrote: > On Thu, Oct 17, 2019 at 4:43 PM Will Deacon <w...@kernel.org> wrote: > > > > Note that the workaround code ended up being based on -rc2, so I had a > > bit of a faff trying to generate the right diffstat for this pull request > > after merging that branch into our fixes branch based on -rc1. In the end > > I had to emulate the pull locally because I couldn't figure out how to > > drive request-pull correctly despite the shortlog being correct. I'd love > > to know what I should've done instead. > > You did the right thing. > > When there are multiple merge bases, a regular "git diff" doesn't work > since it's fundamentally about two end-points (well, it _can_ work > almost by mistake, but doesn't work in the general case). So the only > way to get a "proper" diff is to do a merge and then diff the result. > > That said, I also accept the output of "git diff" which will then have > a lot of noise from all the _other_ work done between the two merge > bases. I can figure out what happened, and do my own two-endpoint diff > and see what happened, and still se that "yes, that's what the pull > request meant, and that's why the diffstat is garbage". > > What you did is the "good quality" pull request, though.
Thanks, that's helpful to know for next time. I guess I'm most surprised by the discrepancy between the shortlog and the diffstat, whereas I intuitively expected them to be generated in the same way. Will
