LWN: Comments on "LKML archives on lore.kernel.org"

LKML archives on lore.kernel.org

nix — Fri, 06 Jul 2018 10:27:22 +0000

As that page notes, grafts are obsolete now. You want to use 'git replace' instead.

LKML archives on lore.kernel.org

neilbrown — Sun, 24 Jun 2018 22:20:58 +0000

> See https://archive.org/details/git-history-of-linux for an example of it applied to the Linux kernel history.

or closer to home:

https://git.kernel.org/pub/scm/linux/kernel/git/history/h...

LKML archives on lore.kernel.org

cesarb — Sun, 24 Jun 2018 14:24:03 +0000

You can use grafts (https://git.wiki.kernel.org/index.php/GraftPoint) to artificially add extra parents to a commit, so the initial commit of one history appears to have the final commit of the other history as a parent. The SHAs of the commits are unchanged. See https://archive.org/details/git-history-of-linux for an example of it applied to the Linux kernel history.

LKML archives on lore.kernel.org

epa — Sun, 24 Jun 2018 13:46:57 +0000

A pity that git doesn't allow you to splice together two histories keeping the existing SHAs valid. So all commits after the splice point would have two SHAs and could be referred to by either. Then the Git kernel history could be spliced on to the git repository of Bitkeeper-era history. (Allowing multiple SHAs for a single commit might also help migrate away from SHA-1, when that day comes.)

LKML archives on lore.kernel.org

cesarb — Sat, 23 Jun 2018 11:33:35 +0000

Yes, there's already a repository somewhere with every Bitkeeper commit of the Linux kernel history converted into the equivalent git commit. For the pre-Bitkeeper history, AFAIK the best we have is a repository which converts each pre-patch (which normally contains several unrelated changes) into a single git commit.

LKML archives on lore.kernel.org

grawity — Sat, 23 Jun 2018 09:11:07 +0000

By "pre-DVCS" you also mean pre-BitKeeper, I assume?

LKML archives on lore.kernel.org

cesarb — Fri, 22 Jun 2018 21:32:40 +0000

This new archive has things I couldn't find anywhere else, like a couple of patches I wrote to the Linux kernel nearly two decades ago (amusingly, the second one can still be found on Linux 4.17, unchanged except for the indentation level).

Which leads me to think: this archive could perhaps be used to fill more of the gaps in the pre-DVCS Linux history. It might be possible to crawl this archive to partially reconstruct the set of original patches which were applied to each kernel (pre-)release, and construct a git repository from that so that "git blame" could point to the original author (and most importantly, patch description) of such ancient code.