Template files location

Template files location

[Petr Baudis]

Dear diary, on Fri, Aug 05, 2005 at 06:37:11AM CEST, I got a letter
where Junio C Hamano <junkio@cox.net> told me that...
> We install files in $(HOME)/etc/git-core/templates/ directory.
> 
> In the standard binary distribution scheme, it should probably
> go to either /usr/share or /etc; the former if the distribution
> prefers static boilerplate files in /usr/share, latter if we
> consider them site-specific conffiles.  I just discussed this
> privately with a Debian person on the list today, and we are in
> favor of installing it under /usr/share/git-core/templates,
> because we do not expect the contents of the standard templates
> directory will not be a per-site customization [*1*].
> 
> In any case, this means that the make drivers like spec.in and
> debian/rules would need to override some make variables when
> they invoke the main Makefile.

Any reason why that's not the default destination then?

Or you can try /etc/git-core/ and fall back to /usr/share/git-core :-)

-- 
				Petr "Pasky" Baudis
Stuff: http://pasky.or.cz/
If you want the holes in your knowledge showing up try teaching
someone.  -- Alan Cox

Re: Template files location

[Junio C Hamano]

Petr Baudis <pasky@ucw.cz> writes:

> Any reason why that's not the default destination then?

Inertia, installing things under $HOME, is why it is not the
default.  It may be that the project is mature enough that it is
time to move away from default installation in HOME, IOW to
change the Makefile to say prefix=/usr/local (or prefix=/usr)
to match practices of other projects [*1*].

> Or you can try /etc/git-core/ and fall back to /usr/share/git-core :-)

If the template files are to become something that always have
to exist, /etc first and then falling back on /usr/share would
make a lot of sense.  But as Johannes Schindelin correctly
argued against the "Use the template mechanism to set up refs/
hierarchy as well." patch [*2*], I think git-init-db should work
when there is no template directory.  In other words, its
primary purpose is to help local project administrators ensure
newly created repositories have hooks and probably info/exclude
that they recommend to the project members.

The reason to have a sample one shipped as part of the core
package is just to help newbies --- they would get a boilerplate
hooks/update that explains how they can set it up when they do
git-init-db even when they do not have their own customized set
of templates yet.  For this kind of use, I do not think one
default falling back to another is needed.

Come to think of it, GIT_TEMPLATE_DIRECTORY environment variable
does not make _ANY_ sense; I am an idiot.  It is used only by
git-init-db and the reason to have it is to override the
default.  It should become the command line parameter of it.
I'll fix up this breakage soonish.


[Footnote]

*1* This would probably break Linus, myself and others ---
everybody has to say "make prefix=$HOME", so I do not think I am
actually going to do it any time soon, if ever.

Having prefix=/usr/local as default only helps people who are
installing system-wide from the source, and nobody else.  People
who are writing spec.in and/or debian/rules need to override it
to prefix=/usr anyway, and it forces people who are installing
to their home to say prefix=$HOME/.  I suspect it is an inertia
from the good old days when nobody used binary distributions.

*2* And I am sure many others shared the same objection but did
not even bother to say anything because what Johannes said made
a lot of sense and what the patch did was obviously wrong.

Re: Template files location

[A Large Angry SCM]

Junio C Hamano wrote:
...
> If the template files are to become something that always have
> to exist, /etc first and then falling back on /usr/share would
> make a lot of sense.  But as Johannes Schindelin correctly
> argued against the "Use the template mechanism to set up refs/
> hierarchy as well." patch [*2*], I think git-init-db should work
> when there is no template directory.  In other words, its
> primary purpose is to help local project administrators ensure
> newly created repositories have hooks and probably info/exclude
> that they recommend to the project members.

So templates are project specific.

> The reason to have a sample one shipped as part of the core
> package is just to help newbies --- they would get a boilerplate
> hooks/update that explains how they can set it up when they do
> git-init-db even when they do not have their own customized set
> of templates yet.  For this kind of use, I do not think one
> default falling back to another is needed.

Shipping samples with the plumbing makes sense, especially when the 
documentation in is insufficient. But installing the samples as part of 
the default install process seems less than desirable. Or, alternately, 
install the samples in with the documentation as (non-executable) examples.

...
> *1* This would probably break Linus, myself and others ---
> everybody has to say "make prefix=$HOME", so I do not think I am
> actually going to do it any time soon, if ever.
> 
> Having prefix=/usr/local as default only helps people who are
> installing system-wide from the source, and nobody else.  People
> who are writing spec.in and/or debian/rules need to override it
> to prefix=/usr anyway, and it forces people who are installing
> to their home to say prefix=$HOME/.  I suspect it is an inertia
> from the good old days when nobody used binary distributions.

Rather than changing the default install location in such a way as to 
make half the user unhappy, make everybody (un)happy by removing the 
default and forcing it to be specified when make is executed.

Re: Template files location

[Petr Baudis]

Dear diary, on Fri, Aug 05, 2005 at 07:33:02PM CEST, I got a letter
where Junio C Hamano <junkio@cox.net> told me that...
> *2* And I am sure many others shared the same objection but did
> not even bother to say anything because what Johannes said made
> a lot of sense and what the patch did was obviously wrong.

Me too, actually. I just didn't step out to see the big picture when
writing the grandparent. Now I agree with everything you said.

-- 
				Petr "Pasky" Baudis
Stuff: http://pasky.or.cz/
If you want the holes in your knowledge showing up try teaching
someone.  -- Alan Cox

Re: Template files location

[Matthias Urlichs]

Hi, A Large Angry SCM wrote:

> Rather than changing the default install location in such a way as to make
> half the user unhappy, make everybody (un)happy by removing the default
> and forcing it to be specified when make is executed.

Lots of stuff installs to /usr/local if not stated otherwise.
I don't see why git should be different.

Why make *everybody* unhappy if we can avoid it?

-- 
Matthias Urlichs   |   {M:U} IT Design @ m-u-it.de   |  smurf@smurf.noris.de
Disclaimer: The quote was selected randomly. Really. | http://smurf.noris.de
 - -
Grain grows best in shit
		-- U. K. LeGuin