this demo goes with 'gittalk.notes' in my 'git-notes' repo, which can be seen in 'rendered' form at http://sitaramc.github.com/other-stuff/gittalk.html
C
Fetching latest commit…
Cannot retrieve the latest commit at this time.
Permalink
Failed to load latest commit information.
external
sampleproj.save
.gitignore
01-simple
01a-simple
02-clone
02helper
03-normal-cycle
04-short-branch
05-long-branch
06-merge-info
07-log
08-show
09-refactor+bug-fix
10-cherry-pick
11-rebase-interactive
12-parallel-branches
13-serial-branches
99-clean
README.md
demo

README.md

notes on the demos in my "why git" talk

This is a quick rundown on the demos that form part of 'gittalk.notes' in my 'git-notes' repo, (a presentation which can be seen in 'rendered' form at http://sitaramc.github.com/other-stuff/gittalk.html).

The demos are a kludge, make no mistake. Their primary purpose is to avoid actually typing in a series of commands -- they save time as well as make sure I don't forget a step in the demo.

This doesn't apply to the GUI parts -- I can't script them so easily, but that's fine; the level of my audience is usually such that they would want to see this being done properly. YMMV.

Prerequisites

In the (currently empty) directory called "external", get a clone of cobbler:

cd external
git clone git://git.fedorahosted.org/hosted/cobbler

I wanted a real-life project that wasn't too big in size or too complex in activity, and cobbler was the first one I explored that fit the bill. There is nothing special about cobbler otherwise. However, a lot of the actual demos now depend on cobbler (for example, some pieces have hardcoded SHAs; they allow me to directly jump to bits of the commit tree that are useful in pointing out various aspects of git)

So if you want to use some other project, you certainly can -- but you'll have to change a lot of things in the demo to fit...!

Structure

Driver

./demo is the main program you run each time you want to do a demo, like so:

./demo 07

Each demo is scripted in a file with a name like 07-log; when you type ./demo XX, a file whose name is XX-* is picked up (there should only be one).

Scripts

Each script is just a series of shell commands. If the command starts with a ^ (hat), it is silently run by ./demo. Otherwise, the command is printed to screen -- so you can point out to the audience what command you are "running" -- and when you hit enter, it actually runs the command.

You can type n before you hit enter on one of those "prompts" and that command won't be run. However, although I put that feature in, I never actually used it in all the times I have given this talk.

Special scripts

Some scripts call other ones to prep the repo for that particular demo, plus there's a helper function (02helper -- note the lack of a hyphen in its name) that other scripts often call. This works behind the scenes; it copies external/cobbler as cobbler and removes the external branches to avoid confusing people when we show them gitk or something.

You can actually run/re-run the demos in any order -- there is no need to stick to the sequence.

In addition, running ./demo 99 cleans up all temporary stuff and returns you to a clean slate.

Details of each demo

01-simple

A very basic starter demo starting with just 3 files (hello.c, Makefile, README.txt), git init, etc.

Also shows the use of gitignore.

When vim .gitignore runs, type the word "hello" into the editor and save it. When the commit editor comes up, point out the contents of the editor (in the "comments"), type up a decent commit message, and save.

ALTERNATIVE SCRIPT: ./demo 01a does the same thing, but replaces the vim with an echo hello > .gitignore so you don't even have to do that. It also does the commit the same way, using git commit -m ...

02-clone

Calls 02helper (see above). This demo shows the git clone command, and a dummy git pull and git push so you can tell them what they look like when there are no changes to be propagated either way.

03-normal-cycle

This starts up to vim sessions -- the first to change an existing file, the second to create a new one. It shows git diff, git status, git add filename, etc.

It then starts up gvim (you may want to change this is you do not like/use gvim), then git gui, git status, git show, and finally ends with a git push.

Make some simple changes in each file that comes up in vim/gvim. When git gui runs, use it to make a commit (show them staging by clicking on the icon next to the file, type in some comment text in the box, click the commmit button)

04-short-branch

This just shows a short branch in gitk so they get a feel for what it looks like. The SHA is hardcoded in the script, so if you use some other repo than cobbler just pick an appropriate one and change it.

05-long-branch

This shows a longer branch using gitk. In addition, there is a second call to gitk using --date-order -- it presents the same long branch in a subtly different way that is sometimes useful to demonstrate. As you page down the tree in both modes you'll see the difference.

06-merge-info

This opens gitk on cobbler at a point where you can see three merges very close to each other; the first two are simple enough, and the third one has a conflict, which was resolved as a part of the merge. You can show, in each case, the sort of information that is available -- the 2 parent SHAs, the fact that the first 2 merges show no actual changes, and the contrast with the 3rd one.

07-log

This just shows various git log options. The two commands showing log "since a certain date" and "...date+time" may or may not show different things based on your time zone. In India, where I live, the former one shows one extra commit.

08-show

Ditto, for git show

09-refactor+bug-fix

This is a nice demo; I usually make sure I have time for this one.

I use two branches, dev1 and dev2 to simulate two separate lines of development by different developers. I also use perl commands to simulate the changes these developers would have normally made using vi, to save time.

Most importantly, the second dev, after making his changes, both renames and moves the file to another directory.

When both these branches are merged into master, git manages to apply both the sets of changes, including the rename, properly.

This is something that no other VCS, not even Mercurial -- which comes closest to git in many ways -- can do.

SPEAKER ACTIONS: This is worth noting down carefully.

  • the demo starts with a gitk &. Do not close this till then end. At each and every command, I switch to this display and do a Ctrl-F5 to demonstrate what the effect of the just-executed command. For example:
    • the effect of a git tag
    • the effect of a git checkout -- point out that the new branch has become bold in the gitk display
    • what a fork looks like
    • what the first "merge" does (it's a fast-forward)
    • (etc)
  • make sure to point out that the perl commands are only a convenience, substituting for vi or some other editor
  • when the first git gui comes up, stage it and commit as normal, then close that window; you don't need it anymore
  • when the second git gui comes up, show them the state of the repo, as seen in gitk, at three points (you must hit Ctrl-F5 each time in gitk):
    • before you do anything in git gui
    • when you have staged just the "new" file
    • after you have staged both of the files -- gitk will suddenly realise (and show) that there is a 97% similarity between those 2 files and will show it as a rename plus some changes. Point out that you did not, at any time, do a git mv or equivalent!
    • after you have committed the second set of changes via git gui
  • once the second git gui has been done, I let the other commands run, (each time showing the gitk output).

10-cherry-pick

This one just shows a cherry-pick. You just have to let the commands run, and then -- using the gitk, show how to cherry-pick a couple of those commits onto master. You can also explain what would happen if you picked them in the wrong order (that is, under what conditions can you expect "problems" by changing the order of two commits -- if commit 2 is in an area of code that was already changed by commit 1, you will have some trouble reversing them).

11-rebase-interactive

A very straightforward demo, with lots of scope for git-evangelism :) Show them that although gitk is constrained to show you commits in a "newest at the top" fashion, the rebase -i has to go "oldest at the top" because that's the sequence in which the commits are going to be applied. Pick a couple, swap one with another, squash one, drop one by deleting it... then save.

Make sure they remember what you did in the editor, hit Ctrl-F5 on gitk, and show them what happened.

12-parallel-branches and 13-serial-branches

These two are meant to show the difference between merging and rebasing, when faced with the common issue of a push being rejected due to being non-fast forward.

You'll have to keep two gitks running in each of these demos, to show graphically what is happening to the commit tree in each of the two repositories (cobbler and cobbler-clone)