is there a truly compelling rationale for .git/info/exclude?

   currently having a discussion with ben straub of "pro git" notoriety,
and he and i seem to agree that there's not much value in registering
ignore patterns in a repo-specific .git/info/exclude file.

   on the one hand, the .gitignore files that come with a repo would
represent (in ben's terminology, which i really like) the "intrinsic"
patterns to be ignored that are related to the basic content of the repo.

   at the other end, users are certainly welcome to add extra patterns
to be ignored, based purely on the way they work -- perhaps based on
their choice of editor, they might want to exclude *.swp files, or
if working on a Mac, ignore .DS_Store, and so on, using a
core.excludesFile setting.

   and in this funny grey area in between, we have .git/info/exclude,
to be used for ... what, exactly? the one argument i've come up with
is the situation where you discover that a repo you've cloned has an
incomplete set of .gitignore patterns, and while you submit a patch
for that to the maintainer, you can temporarily add that pattern
to .git/info/exclude, and as soon as the patch is accepted, you can
toss it.

   but even that isn't a really compelling reason. so what's it for?

rday

rpjday@crashcourse.ca writes:

This is primarily why .git/info/exclude exists. A user who does not use the same set of tools to work on different projects may not be able to use ~/.gitconfig with core.excludesFile pointing at a single place that applies to _all_ repositories the user touches.

Also, core.excludesFile came a lot later than in-project and in-repository exclude list, IIRC.

Don't waste time by seeking a "compelling" reason. A mere "this is the most expedite way to gain convenience" back when something was introduced could be an answer, and it is way too late to complain about such a choice anyway.

On Fri, 6 Oct 2017, Junio C Hamano wrote:

  perfectly respectable answer ... it tells me that, between
.gitignore files and core.excludesFile, there's not much left for
.git/info/exclude to do, except in weird circumstances.

rday

Hi,

Robert P. J. Day wrote:

I use .git/info/exclude in what I don't consider to be weird circumstances.

But I am not motivated to say more than that without knowing what my answer is going to be used for. E.g. is there a part of the gitignore(5) man page where such an explanation would make it less confusing and more useful?

Thanks, Jonathan

On Fri, Oct 06, 2017 at 01:39:16PM -0400, Robert P. J. Day wrote:

A place where I use it is in some Vim package repositories that I have as submodules of my home directory. The author of those repositories, Tim Pope, explicitly does not exclude the helptags output. I simply ignore those files using .git/info/exclude.

Another case is when I install a plugin that lives below a our main product repository at work. I can simply exclude that plugin locally on my system without the need to submit a change for merge. I can later remove those patterns if I like and run git clean -df to clean up.

Hi Robert,

[who I had to cull from the To:/Cc: headers, as my mailer consistently told me that there is no valid DNS record to route mail to rpjday@crashcourse.ca, which *is* weird.]

On Fri, 6 Oct 2017, Robert P. J. Day wrote:

I use .git/info/exclude to keep worktrees in subdirectories of the "main" worktree.

That's not really weird. It's just something few people do, but that's not the same as "weird".

Ciao, Johannes

On Fri, Oct 13, 2017 at 01:18:00AM +0200, Johannes Schindelin wrote:

You are not the only one to mention this, so I did 60 seconds of digging. Turns out that the MX of crashcourse.ca points to a CNAME (mail.crashcourse.ca), which is explicitly forbidden by RFC 2181 (section 10.3). Some MTAs are picky about this and others are not (mine isn't, so I've added Robert back to the cc so he sees this).

-Peff

On Thu, 12 Oct 2017, Jeff King wrote:

  ok, i'll tell my admin about this, thanks.

rday

On Fri, 2017-10-06 at 06:14 -0400, rpjday@crashcourse.ca wrote:

Thanks for asking this question. I have long been in the scenario you
just described above except that I didn't know of .git/info/exclude all
these days. I was longing to find if there was a way to ignore files in
 a repo without touching the .gitignore of that repo . Now I have found
one, the ".git/info/exclude".

Thanks, again.

I use it to ignore stuff in my git-versioned home directory.

Every time I use a new program and it creates a config file or a config directory, it shows up as clutter in magit in my git versioned home directory.

I started with putting the stuff to be ignored in .gitignore, but since I run different stuff on different machines and on different OSes, .gitignore started to contain irrelevant stuff (ignoring a stuff from a program that was run once and then never again, ignoring stuff on one machine that maybe should not be ignored on a different machine), and then I figured it was much simpler to just ignore stuff repo-locally in .git/info/exclude