Beat Bolli <[email protected]> writes: > On 24.07.18 20:22, Junio C Hamano wrote: > >>> This was already fixed (differently) in >>> <[email protected]>. >> >> Thanks for saving me from having to dig the list archive myself. >> Yes, it is already applied to the tip of the topic that originally >> caused the breakage. >> > Just a general question: > > Is it OK to refer to patches on pu with the Message-ID, or would you > prefer the commit hash? The hash changes whenever you recreate pu, > doesn't it?
Either is fine in practice. The commits themselves on a topic branch that is not yet in 'next' usually stay the same once the tip of 'pu' that contains them gets published. Even though I often use "git rebase -i", "git commit --amend", etc. to fix up posted patches while turning them into commits on topic branches, I usually stop doing so once I push out day's integration result. Until a new version of the series is posted to replace them on the topic branch, that is. But at that point we are talking about new patches with different message-ids that got turned into different commit objects, so either commit object name or message id that refer to older iteration would still name the same old version, and new names would refer to the same new version.
