[ofiwg] linux coding style question, or what is the libfabric interpretation of "resembles"

Blocksome, Michael michael.blocksome at intel.com
Wed Mar 11 09:20:04 PDT 2015

> From: Jeff Squyres (jsquyres) [mailto:jsquyres at cisco.com]
> On Mar 11, 2015, at 11:42 AM, Blocksome, Michael
> <michael.blocksome at intel.com> wrote:
> >
> > I think it can be done safely if the threshold for giving up is low. For
> > example, only link to the ofiwg pre-commit hook if there is no pre-existing
> > hook.  Otherwise, dump a warning/info message and do nothing.
> But that's undesirable because:
> 1. I already have a pre-commit hook, but I'll run the libfabric hook at the end
> of mine, and I therefore don't need a suggestion to install the libfabric one.

Sure. You are an advanced user of git. Is that a problem to see help text intended
for others? The help text would be targeted to those who are not advanced
users of git, and yet have some pre-existing hooks for whatever reason. Novice
users will not have the hook. Advanced users will incorporate the hook themselves.

> 2. After it installs the pre-commit hook the first time, it's going to offer a
> suggestion of installing it (again) every other time because it will detect that a
> hook is now there.

The autogen.sh script (or whatever) can check if the pre-existing hook is a link
into the ofi source and then not offer the help text.

> > The user will just have to be surprised when their code is rejected at the
> > server, and we can provide the same help text at that time as a polite
> > "suggestion". :)
> Keep in mind that we can't reject commits at the server -- github doesn't
> offer the capability of a pre-commit hook (only a post-commit hook).

Right .. but doesn't the development process require the maintainer to merge
and/or perform pull requests before code gets in to master? Only a select few
can push to master, right?

More information about the ofiwg mailing list