I've pushed 2.10pre to [email protected]:reposurgeon/reposurgeon.git
and there's a new version of the conversion metadata at [email protected]:reposurgeon/nut-conversion.git I'm not entirely happy with the branch link deduction, but I think this is the best we're going to do without manual intervention to break some harmless but spurious parent links. The underlying problem is twofold: (1) The repo history is cluttered with odd little single-file copies, probably mostly generated by cvs2svn. These show up in gitk as merge bubbles with no commits on one side. The smallest such glitch looks like this in the commit graph: | o |\ | \ | o | / |/ o | (2) There are a couple of branch creations that were done wrong, as a conventional (non-Subversion) directory copy followed by an add. These check out OK but are missing the branch link information perovided by a directory copy operation. Because (1) is true, I cannot assume that every file copy implies a branch link. Because (2) is true, I cannot only use directory copies to impute branch structure - sometimes, I have to notice file copies by comparing hashes and deduce that a branch creation was intended but fumbled. The compromise I'm using is to ignore *single* file copies in computing branch structure - that is, if I detect more than one per changeset I treat that changeset as a branch link. This rule pops almost all spurious merge bubbles, but not all. Charles, unless you spot a place where my branch analysis has gone wrong, I think that part is done. All that's left to do is make the code for manually creating clean branch merges work. -- <a href="http://www.catb.org/~esr/";>Eric S. Raymond</a> The kind of charity you can force out of people nourishes about as much as the kind of love you can buy --- and spreads even nastier diseases. _______________________________________________ Nut-upsdev mailing list [email protected] http://lists.alioth.debian.org/cgi-bin/mailman/listinfo/nut-upsdev
