[merge] global (per user) ignores and other cleanups

[Wayne Davison]

On Fri, Jul 07, 2006 at 02:52:26PM -0500, John Arbash Meinel wrote:
> I'm just more concerned about the opposite. Where a project should be
> self-contained to ignore the right things. And *.o and *.py[co] are
> very likely to end up in a user's default ignores because they get
> tired of adding it all the time.

Yes, some folks could end up doing the CVS-style thing and insert a
global ignore of *.o into their per-user ignore file, but I think that
this is an argument in favor of educating users to not do this rather
than an argument that a per-user ignore file is not a good idea.  For
instance, my ignore file has "x.patch" in it, and I don't think that
any project should be adding my own personal temporary filename choices
to their .bzrignore file.  Also, I think that a deficient .bzrignore
file is an easily solved problem as more people start to use a project:
this is because the missing rules are easily spotted, and patches to
fix things will be forthcoming from the new users (or, if required,
maintained locally).

One potential means of helping people to get their ignore rules into
the right spot might be to create a default ~/.bazaar/ignore file for
those that lack one, and include some comments at the top:

 # This file should be used for adding personal ignore rules and for
 # ignoring more "ephemeral" files that are not created by the build
 # process for a project.  Rules for (non-committed) generated files
 # (such as *.o) should be placed in each project's .bzrignore file.

> The other possibility would be that if ~/.bazaar/ignore *doesn't* exist,
> it uses the builtin list of default ignores. Which is more of a
> compatible upgrade path than our current system.

My original patch did that (used the default ignore list only when the
ignore file was missing).  If this is done, the old ignore list should
be trimmed down to only mention "ephemeral" files.

Thanks for your much-improved patch.

..wayne..