A weird commit from 1970

[mgol]

GitHub detects one Git commit by @skomski from... 1970. This is breaking the contributors page. It was OK few months ago. Screenshot:

[rvagg]

Something new for node-accept-pull-request to test for. If the commit is close to HEAD I'd be inclined to support a rebase and fix.

[jbergstroem]

We briefly discussed this here #2402 (comment) where you also can find the relevant commit.

[thefourtheye]

@rvagg it was first spotted in #2402 (comment). I am not sure if we can fix it now :(

[thefourtheye]

@jbergstroem speak of the timing :D

[mscdex]

Too bad when you highlight a span of time in that github timeline, it doesn't zoom that portion to fit the container.

[rvagg]

Here's what's on top of that commit so far:

2015-09-04 14:48:13 -0700 - 2015-09-05 10:09:16 -0400 - deps: create .npmrc during npm tests
2015-09-04 14:47:14 -0700 - 2015-09-05 10:08:11 -0400 - deps: upgrade to npm 2.14.2
2015-08-11 00:48:49 -0700 - 2015-09-04 20:28:06 -0700 - deps: backport 75e43a6 from v8 upstream (again)
2015-09-04 15:41:21 +1000 - 2015-09-05 13:26:58 +1000 - build: fix .pkg creation tooling
2015-09-03 16:22:57 +1000 - 2015-09-05 11:51:02 +1000 - doc: add TSC meeting minutes 2015-09-02
2015-09-04 12:48:20 +0200 - 2015-09-04 20:42:59 +0200 - doc: update environment vars in manpage and --help
2015-09-03 10:10:30 +0200 - 2015-09-04 11:28:40 -0600 - src: use standard conform snprintf on windows
2015-09-03 10:10:29 +0200 - 2015-09-04 11:27:07 -0600 - src: fix buffer overflow for long exception lines
2015-09-02 14:54:24 +0530 - 2015-09-04 17:12:53 +0200 - doc,test: enable recursive file watching in Windows
2015-09-01 10:51:17 +0200 - 2015-09-04 08:06:52 +0200 - buffer: SlowBuffer only accept valid numeric values
2015-09-04 12:25:31 +1000 - 2015-09-04 12:34:44 +1000 - test: fix use of `common` before required
2015-08-31 17:10:53 -0700 - 2015-09-03 16:38:16 -0700 - deps: upgrade V8 to 4.5.103.30
2015-08-28 09:02:32 +0200 - 2015-09-03 16:38:16 -0700 - src: re-enable fast math on arm
2015-08-23 12:53:05 -0700 - 2015-09-03 16:38:16 -0700 - src: enable vector ics on arm again
2015-07-18 11:34:16 +0200 - 2015-09-03 16:38:15 -0700 - src: replace usage of v8::Handle with v8::Local v8::Handle is deprecated: https://codereview.chromium.org/1224623004
2015-07-01 23:47:37 +0200 - 2015-09-03 16:38:15 -0700 - src: enable v8 deprecation warnings and fix them
2015-08-23 07:24:54 -0700 - 2015-09-03 16:38:15 -0700 - test: fix test-repl-tab-complete.js for V8 4.5
2015-07-07 15:27:14 -0700 - 2015-09-03 16:38:15 -0700 - contextify: ignore getters during initialization
2015-08-23 06:28:57 -0700 - 2015-09-03 16:38:15 -0700 - src: apply debug force load fixups from 41e63fb
2015-08-23 06:09:40 -0700 - 2015-09-03 16:38:15 -0700 - deps: upgrade V8 to 4.5.103.24
2015-09-02 12:23:49 -0400 - 2015-09-03 17:32:51 -0400 - events,lib: don't require EE#listenerCount()
2015-08-14 10:07:18 +0200 - 2015-09-03 13:52:14 -0700 - build: add --enable-asan with builtin leakcheck
2015-09-02 14:16:24 -0700 - 2015-09-03 10:51:30 -0700 - child_process: check execFile and fork args
2015-08-28 19:56:22 -0700 - 2015-09-03 10:39:12 -0700 - test: refactor to eliminate flaky test
2015-08-28 16:58:18 -0400 - 2015-09-03 12:11:45 -0400 - doc: update url doc to account for escaping
2015-07-25 16:09:59 +1000 - 2015-09-03 06:29:27 -0400 - doc: reorder collaborators by their usernames
2015-09-02 11:30:00 +0200 - 2015-09-03 00:27:31 -0400 - test: mark eval_messages as flaky
2015-08-31 00:49:34 +0200 - 2015-09-02 19:39:35 -0400 - child_process: add callback parameter to .send()
1970-08-16 16:09:02 +0200 - 2015-09-02 18:39:39 -0400 - src: fix memory leak in ExternString

I personally have no qualms about rebasing a fix on that commit which would obviously mean changing hashes from there onwards, but I'm generally pretty relaxed wrt my git logs. @nodejs/tsc anyone feel strongly about not wanting to mess up the HEAD of the active branches (this would involve v3.x and v4.x too)?

[silverwind]

We must save our history!

[Fishrock123]

Uh, +1 fixing the commit.

[rvagg]

OK everyone, @nodejs/collaborators, I've applied the heavy hand after not getting any objection here and pinging various TSC members via other channels. So heads-up, you're going to have to reset your branches to the new remote branches.

for i ingit branch -r | grep origin; do echo $i; git log --oneline $i | grep 'src: fix memory leak in ExternString'; done showed that the commit had made it's way to the following branches:

origin/master
origin/v3.3.1-release
origin/v3.x
origin/v4.0.0-rc
origin/v4.x

The v3.3.1-release branch is just a release proposal and the only tag we're tampering with here is one I made for v4.0.0-rc.1 that sits on the v4.0.0-rc branch which has patches I've prematurely landed just to get an RC out.

I've amended the commit on each of these branches, using git commit --amend --date "Sun, 16 Aug 2015 16:09:02 +0200" which I believe is the correct original author date (the GitHub API tells me the PR was made not long after this time so I just adjusted the year portion of the original date). This means that downstream commits, listed above, have been rebased and had their hashes updated on each of the impacted branches. The changes have all been force pushed back up. This will obviously cause a tiny headache for those who have recent copies on your local machines. git reset origin/branch --hard I think is probably the easiest way to do this, or simply delete your local branch and checkout the nodejs/node version again. I've also recreated the v4.0.0-rc.1 tag and force pushed that.

On the brighter side, https://github.com/nodejs/node/graphs/contributors is fixed and the history is sane once again.

[silverwind]

This should fix your local master (repeat for each branch, local changes will be lost):

git reset --hard 617ee324~1 && git fetch origin && git merge --ff-only origin/master