#!/bin/sh APP_NAME=https-everywhere # builds a .xpi from the git repository, placing the .xpi in the root # of the repository. # invoke with no arguments to pull a prerelease of the current # development point. # invoke with the literal argument "uncommitted" to build from the # current src directory. # invoke with a tag name to build a specific branch or tag. # e.g.: # ./makexpi.sh 0.2.3.development.2 # or just: # ./makexpi.sh # BUGS: if you have a branch or tagged named "uncommitted" then this # is kind of ambiguous. cd "$(dirname $0)" if [ -n "$1" ] && [ "$1" != "uncommitted" ]; then VERSION="$1" TARG="$1" else VERSION="$(grep em:version src/install.rdf | sed -e 's/[<>]/ /g' | cut -f3)~pre" TARG=HEAD if [ "$1" != "uncommitted" ] && [ -n "$(git status src -s)" ] ; then printf >&2 "\ WARNING: There are uncommitted changes in your current repostitory. WARNING: These changes will not be included in the generated .xpi WARNING: Run 'git status' for information about the uncommitted changes. WARNING: Or, use 'makexpi.sh uncommitted' to include them in the build. " fi fi XPI_NAME="pkg/$APP_NAME-$VERSION.xpi" [ -d pkg ] || mkdir pkg cd "src" if [ "$1" = "uncommitted" ]; then printf >&2 "WARNING: using zip instead of git archive to build .xpi\n" CHANGES="$(git status . -s)" if [ -n "$CHANGES" ]; then printf >&2 "WARNING: uncommitted changes were included:\n%s\n" "$CHANGES" fi # FIXME: is it really acceptable to reuse .gitignore to specify # include patterns for /usr/bin/zip? It seems to work for our # current patterns (2010-11-09) zip -X -q -9r "../$XPI_NAME" . "-x@../.gitignore" else git archive --format=zip -9 "$TARG" . > "../$XPI_NAME" fi ret="$?" if [ "$ret" != 0 ]; then rm -f "../$XPI_NAME" exit "$?" else printf >&2 "Created %s\n" "$XPI_NAME" fi