[Buildroot] [RFC PATCH 5/9] support/scripts: add create-relocation-script for toolchain relocation

[Arnout Vandecappelle]

On 03-03-17 15:18, Wolfgang Grandegger wrote:
> It will create the script "relocate-toolchain.sh" in $HOST_DIR/usr/
> allowing to adjust the path to the toolchain directory in all text
> files after it has been moved to a new location.
> 
> Signed-off-by: Wolfgang Grandegger <wg at grandegger.com>
> ---
>  support/scripts/create-relocation-script | 72 ++++++++++++++++++++++++++++++++
>  1 file changed, 72 insertions(+)
>  create mode 100755 support/scripts/create-relocation-script
> 
> diff --git a/support/scripts/create-relocation-script b/support/scripts/create-relocation-script
> new file mode 100755
> index 0000000..0ceaeb2
> --- /dev/null
> +++ b/support/scripts/create-relocation-script
> @@ -0,0 +1,72 @@
> +#!/bin/sh
> +
> +# Creates the script "relocate-toolchain.sh" and the "location" file in
> +# the buildroot toolchain (host/usr). After copying that toolchain tree
> +# to a new location, the script in the top directory should be executed
> +# to relocate the toolchain. It actually will replace the string of the
> +# old location with the new one in *all* text files.
> +
> +if [ "$#" -ne 1 -o ! -d "$1" ]; then
> +    echo "Usage: $0 <buildroot-host-directory-path>"
> +    exit 1
> +fi
> +
> +# The toolchain is in buildroot's "host/usr"
> +TOOLCHAINDIR="$1/usr"
> +# We create the script in the top directory of the toolchain
> +SCRIPTFILE="${TOOLCHAINDIR}/relocate-toolchain.sh"

 I think installing the script in host/usr/bin would be more appropriate.

> +# File holding the location of the buildroot toolchain
> +LOCATIONFILE="share/buildroot/toolchain-location"
> +
> +echo "Creating ${SCRIPTFILE} for toolchain relocation ..."
> +
> +cat << EOF > "${SCRIPTFILE}"

 Instead of using cat to instantiate the template, it's easier to understand
what happens by creating a template file
(support/misc/relocate-toolchain.sh.in) and using sed to instantiate the
variables that need to be instantiated. Such a sed command can typically be
called directly from Makefile, without helper script.

 However, as far as I can see, the only thing that is instantiated in this
script is LOCATIONFILE. But that variable is actually constant, so it could be
coded directly in the script. So why not just $(INSTALL) the script itself, like
is done for e.g. support/misc/target-dir-warning.txt ?


> +#!/bin/sh
> +#
> +# Automatically generated by $0: don't edit
> +#
> +if [ "\$#" -ne 0 ]; then
> +    echo "Run this script to relocate the buildroot toolchain at that location"
> +    exit 1
> +fi
> +
> +FILEPATH="\$(readlink -f \$0)"
> +NEWPATH="\$(dirname \${FILEPATH})"
> +
> +cd \${NEWPATH}
> +if [ ! -r ./${LOCATIONFILE} ]; then
> +    echo "Previous location of the buildroot toolchain not found!"
> +    exit 1
> +fi
> +OLDPATH="\$(cat ./${LOCATIONFILE})"
> +
> +if [ \${NEWPATH} = \${OLDPATH} ]; then
> +    echo "This buildroot toolchain has already been relocated!"
> +    exit 0
> +fi
> +
> +echo "Relocating the buildroot toolchain from \${OLDPATH} to \${NEWPATH} ..."
> +
> +# Make sure file uses the right language
> +export LC_ALL=C
> +# Replace the old path with the new one in all text files
> +for FILE in \$(grep -r  "\${OLDPATH}"  . -l ) ; do

 We typically collect options before the regex, so grep -r -l ...

 I think that there are a few packages that might install something with spaces
(though probably not in $HOST_DIR), so find -exec is probably more appropriate.
Unfortunately, that doesn't allow you to do a test with a pipe however so a
different solution would be needed to support files with spaces.

> +    if [ ! -h \${FILE} ] && [ -n  "\$(file -b --mime-type \${FILE} | grep '^text/' )" ] && [ "\${FILE}" != "./${LOCATIONFILE}" ]
> +    then
> +        sed -i s,"\${OLDPATH}",\${NEWPATH},g \${FILE}
> +    fi
> +done
> +
> +# Finally, we update the location file itself
> +sed -i s,"\${OLDPATH}",\${NEWPATH},g ./${LOCATIONFILE}

 Why do this separately? It works within the loop above as well, doesn't it?
> +
> +# Check if the path substitution did work properly
> +if [ "\${NEWPATH}" != "\$(cat ./${LOCATIONFILE})" ]; then
> +    echo "Something went wrong with substituting the path!"
> +    echo "Please choose another location for your toolchain!"
> +fi

 Is there any reason why this error handling is needed? The only thing I can
imagine is concurrently running two instances of the script, but otherwise I see
no way that the sed could fail...

 Regards,
 Arnout

> +EOF
> +chmod +x "${SCRIPTFILE}"
> +mkdir -p $(dirname "$1/usr/${LOCATIONFILE}")
> +# Finally write the toolchain location
> +echo "$TOOLCHAINDIR" > "${TOOLCHAINDIR}/${LOCATIONFILE}"
> 

-- 
Arnout Vandecappelle                          arnout at mind be
Senior Embedded Software Architect            +32-16-286500
Essensium/Mind                                http://www.mind.be
G.Geenslaan 9, 3001 Leuven, Belgium           BE 872 984 063 RPR Leuven
LinkedIn profile: http://www.linkedin.com/in/arnoutvandecappelle
GPG fingerprint:  7493 020B C7E3 8618 8DEC 222C 82EB F404 F9AC 0DDF