forgot to reply to list On Sun, May 11, 2008 at 8:17 PM, Govind Salinas <govind@xxxxxxxxxxxxxxxxx> wrote: > On Sun, May 11, 2008 at 7:28 PM, Junio C Hamano <gitster@xxxxxxxxx> wrote: >> "Govind Salinas" <govind@xxxxxxxxxxxxxxxxx> writes: >> >>> I am writing my wrapper over git bundle and I noticed that the >>> "SPECIFYING REFERENCES" section says that the it will only >>> bundle things that end in something git-show-ref can find. >>> >>> I can probably work around this by silently creating a tag >>> doing the bundle and deleting the tag, but I want to know why >>> this restriction is in there in the first place? If there is a good >>> reason for it then I will probably just add this info to the >>> documentation. >> >> Because bundle is not just a random collection of objects, a tarball of >> your .git/objects/. Instead, it is a (partial) history that leads to a >> particular (set of) versions. >> >> Think of it as what "git fetch $somewhere $that_branch" could give you. >> It is not giving you just a collection of random objects, but you are >> choosing from the endpoint the particular repository ($somewhere) is >> offering you. >> >> When you publish your history to be fetched over the network (or locally >> for that matter), you do not just put bunch of objects there. You give >> branches to mark where the histories end. It's the same deal with >> bundles, and the only difference is the transfer may go over sneakernet. >> >> > > Sure, I understand that. However, I can use a tag to create a bundle > that does not go to an endpoint. I can also advance that branch to > a later commit by whatever mechanism (say committing something) > and then the bundle no longer points to the endpoint, it points to > the middle somewhere. Git, from what I have seen, likes to treat > HEADs as just another commit and it is a bit surprising to see this > particular limitation here. I see this as kind of like git-push where > the person who has the commits is specifying them, and there you > can specify any commit. Although pull/fetch have similar > limitations, so perhaps it is not so surprising. > > If I wanted to share a patch series via bundle and the patches I > wanted went from HEAD~10..HEAD~5 then I *could* checkout -b > HEAD~5 or tag HEAD~5, but I do not see an advantage to doing so. > > I don't really use bundles, so it's not a big deal to me. I just > thought I would ask to make sure I wasn't going to break something. > > Thanks, > Govind. > -- To unsubscribe from this list: send the line "unsubscribe git" in the body of a message to majordomo@xxxxxxxxxxxxxxx More majordomo info at http://vger.kernel.org/majordomo-info.html
