Jan Holesovsky <kendy <at> suse.cz> writes: > But hg-git is a Mercurial extension, and needs git to be the server, doesn't > it? So? The description of hg-git says you can push changesets from a mercurial repo to a git repo keeping their identity. Why shouldnt you do that with an OOo mercurial repo and work on the git repo and push your changeset back as mercurial changesets. If the conversion keeps the identity there shouldnt be a problem. Actually, it should be possible to have a mercurial and a git repo in (daily) sync making the stuff pretty much transparent.
However: - There is no _rational_ reason to do this (however, there are reasons of personal preference/faith ;-) ) - You shouldnt expect any support from RelEng for pulling of such stunts (because could do perfectly well by simply using mercurial) - I havent tested hg-git - no guarantees for its matureness and stability (its not yet distributed with mercurial, but it is actively developed) > Otherwise yes, it would be a killer thing, if both hg and git were usable as > the SCM's for OOo at the same time If you intent to setup an infrastructure to mirror the mercurial-cws's as git-repos making it possible to use git as "enduser" (dev) while changesets are published in mercurial-repos at the same time feel free to do so. However, I have not heard a single argument rationalizing such work (unless you consider "because I can" a valid argument) - up to now, all the "missing features" of mercurial turned out to be just people spending much time learning to use git and little effort to inform themselves about mercurial. I am not a RelEng, but I think one thing has to be clear: Unless some fundamental issue with mercurial turns up (changing the decision for using mercurial as reference SCM), all work intended for contribution should be available in mercurial - with everything else, you are on your own. Just imagine RelEngs work if people would show up with darcs, GNU arch, bzr, old bazaar, SVN, CVS and RCS commits ... Have Fun, Bjoern --------------------------------------------------------------------- To unsubscribe, e-mail: [email protected] For additional commands, e-mail: [email protected]
