[vtk-developers] New git commit hooks

[Bill Lorensen]

Why not keep the TAB check in both places. Even though the local hooks
are optional, developers will soon learn they should enable them
locally. Otherwise pushing will be a PITA.

Bill

On Fri, Apr 23, 2010 at 2:31 PM, Brad King <brad.king at kitware.com> wrote:
> Berk Geveci wrote:
>> I was just describing the conversation I had with Brad. The idea was
>> to have checks only for things that screwed up the history (like
>> merging from next to master) or possibly things that could corrupt
>> people's repositories (I believe an example for that was screwed up
>> line feed). We were not supposed to have commit checks for anything
>> you listed. I somehow don't believe that we are enforcing 79 character
>> or less. If we are, we should stop. Same with trailing white spaces.
>>
>> Brad, can you chime in here please.
>
> The hooks that Robert pointed out at
>
>  http://www.paraview.org/Wiki/VTK/Git#Hooks
>
> can be used by developers locally to avoid creating bad commits in
> the first place.  It does trailing whitespace and commit message checks
> right now.  However, these are *optional* and must be explicitly enabled
> by each developer.  Therefore we should convert the content checks like
> TABs into python tests.
>
> The server has the hooks we discussed:
>
>  - Reject bad author or committer email domains
>  - Reject CRLF, but not trailing whitespace
>  - Reject large blobs (keeps github's http protocol happy)
>  - Reject broken submodule references (push submodules first!)
>
> It looks like I accidentally left the message length enforcement on.
> However, other developers in this thread have expressed a liking for it.
> We should probably remove it on the server so developers don't have to
> rebase and edit long (possibly shared) histories just for that.  The
> local commit check should keep it though.
>
> Comments?
> -Brad
>