Re: Limiting disk usage

From: Piotr Krukowiecki <piotr.krukowiecki@gmail.com>
Date: Oct 30, 2013, 07:44 UTC
Message-ID: <CAA01Cso8J+BBgskoRtxkPRrtM4KJ1O2FhcB4uh+QB2Dr=DZGyQ@mail.gmail.com>
In-Reply-To: <526FCA13.4090902@gmail.com>

On Tue, Oct 29, 2013 at 3:45 PM, Jakub Narębski <jnareb@gmail.com> wrote:

> On 2013-10-29 14:50, Piotr Krukowiecki wrote:
>>
> Lot of code isn't problem, see Linux kernel or GCC, or LibreOffice.org

OK, it depends on how much code is too much for you. I'm not talking about performance of git, just disk usage. In my case I have too much code.

> If they are independent projects, they should get independent repositories;
> you can stitch them back together using git-submodule (or git-subtree).
> reposurgeon can hel you with that.

They are not totally independent projects (nor totally dependent). Normally you want them all, you want to create a branch/tag on all of them, if new directory is created you want to automatically have it too etc. But there are use cases when you need only some of them and then they can be seen as independent .

git-submodule have several disadvantages for my use:
- does not track branches (you have to specify hardcoded SHA1 instead
of a branch name)
- additional complexity / command layer (you have to do things like
"submodule init", "submodule update")
- you can't simply work on all submodules (you can't simply branch,
merge, commit, log, diff, etc in all submodules using one command in
top-level project)

git-subtree looks much better in that regard, but for example it needs an explicit prefix (name of "subproject") on which it operates. So you can't say "update everything" or "commit everything" etc. With hundreds of subprojects you need another layer to automate this...

-- 
Piotr Krukowiecki