[cairo] Changes in web commits

Baz brian.ewins at gmail.com
Thu Jul 5 11:14:52 PDT 2007

On 05/07/07, Carl Worth <cworth at cworth.org> wrote:
> I had glanced very quickly at the git-specific code in ikiwiki and it
> was actually somewhat frightening. It had the appearance of being very
> over-complicated perhaps due to being ported from code that handled a
> version-control system that doesn't provide all of the robustness that
> git provides.

Hmm... this whole rcs backend stuff looks fundamentally wrong, and not
just for git. Its always going to fail if two users are working on the
same branch - and since there's no calls into the rcs backend until
he's already overwritten the current version of the page in CGI.pm, it
is just guaranteed to break at some point.

To be safe, each edit needs to be done into its own clone of the repo,
not its own branch - since he's checking out his 'throwaway branch'
into the current directory, other people are going to be writing into
it too :(

> It might actually make sense to just write new code for creating a git
> commit from a wiki edit from scratch rather than trying to debug
> what's there now.

ways I can see to make it work:
- git clone -s -l master some-unique-name ... etc. Seems a bit
longwinded to commit a change to a single file, and is a git-specific

- construct a patch, apply it with git-am. This seems far easier to
do, and might be a safer approach on all of his backends. Shouldn't
need many forks either? Just git-show, diff, git-am... or am I missing
something here? I don't think theres a command to diff a blob against
an arbitrary file?


More information about the cairo mailing list