Re: How to start well for a special git construction

Thanks for your help.
I've just created my empty new branch for delivery stuff... and
effectively the branch is empty :-D
Any idea on how I can add executable files in this empty branch from
the master branch? Is that possible?
(perhaps I'm using Git in a bad way)

If I want to manage those file with Git, it's just for the purpose to
merge them in another repository. I would have choose to have only one
repository for all things... but basically, the two frameworks are not
maid to be used together. So I thought that using two distinct
repositories was better. Especially for Restfulx code generation which
need a lot of unsalubrious hacks. I don't want to keep them in my
final application...

in any case, thank you very much for your quick answers.


Bruno

2010/3/19, Avery Pennarun [off-list ref]:

On Fri, Mar 19, 2010 at 3:31 PM, bruno le hyaric
[off-list ref] wrote:

quoted

2010/3/18, Avery Pennarun [off-list ref]:

quoted

Do you actually have multiple projects that you're working on, or are
you the maintainer of an application that uses other people's
projects?

I'm starting building a new web application based on 2 existing
frameworks (Spree, RestfulX).

This basic explanation leads me to believe you only really want one
repo (at least once you have it set up): the one for your app.  Other
people already host repos for the other projects.

quoted

I want to use latest upstream of both frameworks. For the moment I
don't have reason to make patches for these, but if I can do that,
this is better. In the other case I can imagine developping patch in
distinct repository in a independant way...

You can embed the contents of the other repositories into yours in one
of three ways that I know of:

- Just copy the files from those libraries into your project's repo.
This is cheap and easy, although it makes it a bit harder to push
patches upstream later, in case you care about that.

- Use git-submodule to import those libraries into your project's
repo.  This is reasonably easy, but makes it hard to patch those
libraries without creating additional repositories, which is
inconvenient.  If you don't need to patch those libraries, it's the
cleanest way.

- Use git-subtree to import those libraries into your project's repo.
This makes it relatively easy to push and pull changes from the
libraries from/to upstream, but can clutter up your 'git log' a bit.

I expect that for your purposes, any of the three options would work fine.

quoted

I plan to distribute the basic work needed by my application.
Basically, I'm building an automated process to get a e-commerce
website with generated Flex (Flash) views. This process is mainly
based on Git/Shell/Rake/Ruby commands.

I won't distribute the source code of my own e-commerce website.

But my problem wasn't about redistribution... RestfulX is a tool for
code generation. It needs some input and generate Flex code as output.
Then I compile Flex code to get a Flash executable (.swf / .swc). I
want to manage everything with versionning, but regarding the merge, I
just want to merge the executable in the destination repository, not
all the stuff used for code generation.

It's very rare with git to commit your output files into a repository.
 You can do it, but it's rare.  Most people just put the numbered
binaries up in a directory on a website somewhere.  Version
controlling binaries any other way doesn't really add anything, since
'git diff' and friends aren't much help, and the output files are (by
definition) possible to reconstruct from the source code, so if you
lose them, it's not a big deal.

If you really want to do it, then you will probably want to at least
put the output files on their own branch (as you suggest) or even into
their own repository (so that people cloning the source repo don't
have to download tons of binary revisions that they won't be using).

Hope this helps.

Have fun,

Avery

On Fri, Mar 19, 2010 at 4:30 PM, bruno le hyaric
[off-list ref] wrote:

Thanks for your help.
I've just created my empty new branch for delivery stuff... and
effectively the branch is empty :-D
Any idea on how I can add executable files in this empty branch from
the master branch? Is that possible?
(perhaps I'm using Git in a bad way)

There are fancy things you can do to create a branch containing only a
single file without checking it out.  It's a bit advanced, but it
looks something like this:

    FILENAME=path/to/file
    BASENAME=$(basename $FILENAME)
    BLOBID=$(cat $FILENAME | git hash-object --stdin -w)
    TREEID=$(printf '100644 blob %s\t%s' $BLOBID $BASENAME | git mktree)
    COMMITID=$(echo Commit Message | git commit-tree $TREEID -p
refs/heads/release-branch)
    git update-ref refs/heads/release-branch $COMMITID

A more straightforward way to do it would be to clone your repository,
and check out the release branch in your clone, then copy the file to
it and commit as you normally would.

Have fun,

Avery