[PATCH] Proposed patch: support for Python functions as .hgignore filters

[Matt Mackall]

On Mon, 2010-06-28 at 15:50 -0500, Steve Borho wrote:
> On Mon, Jun 28, 2010 at 3:43 PM, Matt Mackall <mpm at selenic.com> wrote:
> > On Mon, 2010-06-28 at 22:33 +0200, Mads Kiilerich wrote:
> >> Boris Figovsky wrote, On 06/28/2010 10:07 PM:
> >> > Hello,
> >> > In a project I work on, our build system generated a lot of files in
> >> > the working directory,
> >> > and we wanted Mercurial to ignore them, but the current .hgignore
> >> > syntax is not enough.
> >> > We thought we could use Python functions, such as os.path.islink or
> >> > mymodule.py's is_ignored() func.
> >> > The outcome is the attached patch.
> >>
> >> .hgignore is tracked in the repository, so this patch would allow clever
> >> users to execute any command on your machine if they can trick you to
> >> pull to your repo. (For example by putting "python:os.system" in
> >> .hgignore and creating the file "rm -rf ..".) We don't want that.
> >>
> >> It could work if it used some kind of secure sandboxed version of
> >> python, but that is generally not an option.
> >
> > And Python's extensive introspection makes this basically impossible in
> > CPython.
> 
> Could this be an 'opt-in' feature added by an extension?

Yes, you can add anything via extensions, even gaping security holes.
But we certainly wouldn't ship such a thing.

-- 
Mathematics is the supreme nostalgia of our time.