Re: gnulib bootstrap and git submodules

[Dmitry V. Levin]

Hi,

On Mon, Feb 22, 2010 at 03:58:19PM -0700, Eric Blake wrote:
[...]
> Here's the next three patches I had to add on top of the previously posted
> one.  With this series, I was able to copy the new bootstrap to libvirt
> and still get everything there to work correctly (patch to libvirt list to
> follow soon).  Pushing:

I've got a minor regression with this change.  Here is a bit more context.

[...]
> From a0521b0449635f64694093d08b98ab7d67a28bf4 Mon Sep 17 00:00:00 2001
> From: Eric Blake <address@hidden>
> Date: Mon, 22 Feb 2010 10:59:33 -0700
> Subject: [PATCH 1/3] bootstrap: allow for alternate submodule location
> 
> Libvirt sets submodule.gnulib.path to ".gnulib".  Also, it does
> not need gnulib.mk, but directly uses Makefile.am.
> 
> * build-aux/bootstrap (gnulib_path): New variable; use instead of
> hardcoding submodule location.
[...]
> --- a/build-aux/bootstrap
> +++ b/build-aux/bootstrap
> @@ -390,7 +390,7 @@ fi
> 
>  cleanup_gnulib() {
>    status=$?
> -  rm -fr gnulib
> +  rm -fr "$gnulib_path"
>    exit $status
>  }
> 
> @@ -398,6 +398,8 @@ git_modules_config () {
>    test -f .gitmodules && git config --file .gitmodules "$@"
>  }
> 
> +gnulib_path=`git_modules_config submodule.gnulib.path`
> +

Before this change, git_modules_config was called only if either
no GNULIB_SRCDIR was set or GNULIB_SRCDIR was set to a git working
directory.  Now git_modules_config is called unconditionally, even in case
when its result (gnulib_path variable) is not used later.

A real life example: I've just made a test build of
coreutils-v8.4-63-g9f793de and got the following warning:

$ ./bootstrap --skip-po --gnulib-srcdir=gnulib-v0.0-3591-geb2f221
./bootstrap: Bootstrapping from checked-out coreutils sources...
./bootstrap: line 408: git: command not found

I think it would be OK to move the gnulib_path initialization to code
paths where it is used.


-- 
ldv

pgpLSeTRY7CPH.pgp
Description: PGP signature