Trivia: When did git self-host?

Trivia: When did git self-host?

[Mike Coleman]

Does anyone recall when git first self-hosted?  This doesn't seem to
be present on the GitHistory page on the wiki, and I'm not finding it
elsewhere.

I suppose this could mean different things, but roughly I'm curious
about the earliest point at which all further commits were
human-generated, in real time, as opposed to the machine-generated
commits of some robo-import script.  I'm not seeing this transition in
the git-log, which seems to go continuously back to the very beginning
(which is a little disturbing :-).

Mike

Re: Trivia: When did git self-host?

[Shawn O. Pearce]

Mike Coleman <tutufan@gmail.com> wrote:
> Does anyone recall when git first self-hosted?  This doesn't seem to
> be present on the GitHistory page on the wiki, and I'm not finding it
> elsewhere.
> 
> I suppose this could mean different things, but roughly I'm curious
> about the earliest point at which all further commits were
> human-generated, in real time, as opposed to the machine-generated
> commits of some robo-import script.  I'm not seeing this transition in
> the git-log, which seems to go continuously back to the very beginning
> (which is a little disturbing :-).

I think it was day 1.  One of Linus' goals was to have something
that could track itself, as that meant he was on the right path...
I think he worked about 8 hours or so on Git code before he was
able to actually commit something, and about that point he wanted to
commit.  And that is commit e83c5163316f89bfbde7d9ab23ca2e25604af290.
Yes, the beginning of history.

I've heard that Linus is a very good developer.  ;-)

-- 
Shawn.

Re: Trivia: When did git self-host?

[Johannes Schindelin]

Hi,

On Mon, 26 Feb 2007, Mike Coleman wrote:

> Does anyone recall when git first self-hosted?  This doesn't seem to be 
> present on the GitHistory page on the wiki, and I'm not finding it 
> elsewhere.

Hey, how about adding a trivia page there, with the information you get 
here?

> I suppose this could mean different things, but roughly I'm curious 
> about the earliest point at which all further commits were 
> human-generated, in real time, as opposed to the machine-generated 
> commits of some robo-import script.

The initial commit was v0.99~954, and it was not yet generated with 
git-commit, because it was not there:

$ git show v0.99~954:
tree v0.99~954:

Makefile
README
cache.h
cat-file.c
commit-tree.c
init-db.c
read-cache.c
read-tree.c
show-diff.c
update-cache.c
write-tree.c

Hth,
Dscho

Re: Trivia: When did git self-host?

[Linus Torvalds]

On Mon, 26 Feb 2007, Mike Coleman wrote:
>
> Does anyone recall when git first self-hosted?  This doesn't seem to
> be present on the GitHistory page on the wiki, and I'm not finding it
> elsewhere.

The first commit was already self-hosted. It was done manually (write-tree 
etc by hand), but yes, the first commit really is:

	Thu Apr 7 15:13:13 2005 -0700

and the second one (add copyright notices) was done a few minutes later.

Later that same day I imported "sparse" into a git archive from BK (you 
can get the sparse archive, and see the initial commit:

	commit 3ece2ef7c0a3d5975f65aa09911e1944e4125c45
	Author:     Linus Torvalds <torvalds@home.transmeta.com>
	AuthorDate: Thu Mar 13 12:53:56 2003 -0700
	Commit:     Linus Torvalds <torvalds@ppc970.osdl.org>
	CommitDate: Thu Apr 7 20:59:14 2005 -0700

notice how the commit date is different from the author date - the 
author date was taken from the BK tree that I imported).

The third git commit is committing the changes to git to make that import 
possible ("Thu Apr 7 21:03:28 2005 -0700").

So git was self-hosting since April 7, 2005.

Now, exactly when I started git development (ie how long it took before it 
got to that self-hosting stage), I can't remember. I'd say about two 
weeks, probably.

		Linus

Re: Trivia: When did git self-host?

[Linus Torvalds]

On Mon, 26 Feb 2007, Linus Torvalds wrote:
> 
> The first commit was already self-hosted. It was done manually (write-tree 
> etc by hand), but yes, the first commit really is:
> 
> 	Thu Apr 7 15:13:13 2005 -0700
> 
> and the second one (add copyright notices) was done a few minutes later.
> 
> So git was self-hosting since April 7, 2005.
> 
> Now, exactly when I started git development (ie how long it took before it 
> got to that self-hosting stage), I can't remember. I'd say about two 
> weeks, probably.

Actually, it must have been less than that.

The first version of git was just ~1300 lines of code, and I have reason 
to believe that I started it at or around April 3rd. The reason: I made 
the last BK release on that day, and I also remember aiming for having 
something usable in two weeks. 

And hosting git itself was not that important for me - hosting the kernel 
was. And the first kernel commit was April 16 (with the first merges being 
a few days later). Which meshes with my "two week goal" recollection.

Not that I'd normally be entirely sure I hit any time goals I set, but I 
*am* pretty sure that I decided that in order to be effective, I didn't 
want to do kernel development at the same time as git development, so the 
"April 3" date of the 2.6.12-rc2 release is fairly significant.

			Linus

Re: Trivia: When did git self-host?

[Daniel Barkalow]

On Mon, 26 Feb 2007, Linus Torvalds wrote:

> On Mon, 26 Feb 2007, Linus Torvalds wrote:
> > 
> > The first commit was already self-hosted. It was done manually (write-tree 
> > etc by hand), but yes, the first commit really is:
> > 
> > 	Thu Apr 7 15:13:13 2005 -0700
> > 
> > and the second one (add copyright notices) was done a few minutes later.
> > 
> > So git was self-hosting since April 7, 2005.
> > 
> > Now, exactly when I started git development (ie how long it took before it 
> > got to that self-hosting stage), I can't remember. I'd say about two 
> > weeks, probably.
> 
> Actually, it must have been less than that.
> 
> The first version of git was just ~1300 lines of code, and I have reason 
> to believe that I started it at or around April 3rd. The reason: I made 
> the last BK release on that day, and I also remember aiming for having 
> something usable in two weeks. 

As of April 3rd, you were looking at existing systems, rejecting all of 
them aside from monotone, and finding monotone too slow.

According to the first public mention of git 
(http://lkml.org/lkml/2005/4/8/9), you'd spent two days getting it to the 
point where it was hosting itself and sparse (and doing nothing else), and 
it was ready for other people to start writing things. (That is to say, 
your main task was finding a new SCM, and you were playing with monotone 
as the expected solution, but it was so slow you had time to write git 
while you were waiting for monotone to run)

> And hosting git itself was not that important for me - hosting the kernel 
> was. And the first kernel commit was April 16 (with the first merges being 
> a few days later). Which meshes with my "two week goal" recollection.
> 
> Not that I'd normally be entirely sure I hit any time goals I set, but I 
> *am* pretty sure that I decided that in order to be effective, I didn't 
> want to do kernel development at the same time as git development, so the 
> "April 3" date of the 2.6.12-rc2 release is fairly significant.

You did mention getting 2.6.12-rc2 done, and then spending the ensuing 
calm time doing something about replacing BK, and it sounded like 2 weeks 
was a time limit for getting something in place, even if it was only a 
temporary solution. After two weeks, git was suitable for the immediate 
task of applying patches from other people, and then people started 
figuring out workflows around it.

	-Daniel
*This .sig left intentionally blank*

Re: Trivia: When did git self-host?

[Mike Coleman]

On 2/26/07, Johannes Schindelin <Johannes.Schindelin@gmx.de> wrote:
> Hey, how about adding a trivia page there, with the information you get
> here?

I will add it, to GitTrivia or GitHistory.  Thanks to everyone for the info.

I'm giving a talk about revision control tomorrow, which is what
inspired the question.  I'll be saying that git is very cool.  :-)

Looking back at previous systems, I was surprised at how capable (if
ugly) SCCS appears to have been.  (I got my start on RCS.)  Did you
know it had an "annotate" option (a la 'cvs annotate')?  I didn't.

On another topic, it might amuse you to know that there's yet another
'git' floating around.  It's part of a bioinformatics package called
SEALS.  Though it hasn't been maintained in several years, we have it
installed at work.

Mike

Re: Trivia: When did git self-host?

[Martin Langhoff]

On 2/28/07, Mike Coleman <tutufan@gmail.com> wrote:
> On another topic, it might amuse you to know that there's yet another
> 'git' floating around.  It's part of a bioinformatics package called
> SEALS.  Though it hasn't been maintained in several years, we have it
> installed at work.

There is also GNU Interactive Tools, and some fun flamewars about name
clashes, package conflicts and more... See

  http://lkml.org/lkml/2005/4/13/110
  http://www.gelato.unsw.edu.au/archives/git/0508/7477.html

for the tip of the iceberg.

cheers,


martin