include agent identifier in capability string
[gitweb.git] / t / test-lib.sh
old mode 100755 (executable)
new mode 100644 (file)
index 470a909..9e2b711
 #
 # Copyright (c) 2005 Junio C Hamano
 #
+# This program is free software: you can redistribute it and/or modify
+# it under the terms of the GNU General Public License as published by
+# the Free Software Foundation, either version 2 of the License, or
+# (at your option) any later version.
+#
+# This program is distributed in the hope that it will be useful,
+# but WITHOUT ANY WARRANTY; without even the implied warranty of
+# MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
+# GNU General Public License for more details.
+#
+# You should have received a copy of the GNU General Public License
+# along with this program.  If not, see http://www.gnu.org/licenses/ .
+
+# if --tee was passed, write the output not only to the terminal, but
+# additionally to the file test-results/$BASENAME.out, too.
+case "$GIT_TEST_TEE_STARTED, $* " in
+done,*)
+       # do not redirect again
+       ;;
+*' --tee '*|*' --va'*)
+       mkdir -p test-results
+       BASE=test-results/$(basename "$0" .sh)
+       (GIT_TEST_TEE_STARTED=done ${SHELL-sh} "$0" "$@" 2>&1;
+        echo $? > $BASE.exit) | tee $BASE.out
+       test "$(cat $BASE.exit)" = 0
+       exit
+       ;;
+esac
+
+# Keep the original TERM for say_color
+ORIGINAL_TERM=$TERM
 
 # For repeatability, reset the environment to known value.
 LANG=C
 LC_ALL=C
 PAGER=cat
 TZ=UTC
-export LANG LC_ALL PAGER TZ
+TERM=dumb
+export LANG LC_ALL PAGER TERM TZ
 EDITOR=:
-VISUAL=:
-unset AUTHOR_DATE
-unset AUTHOR_EMAIL
-unset AUTHOR_NAME
-unset COMMIT_AUTHOR_EMAIL
-unset COMMIT_AUTHOR_NAME
-unset GIT_ALTERNATE_OBJECT_DIRECTORIES
-unset GIT_AUTHOR_DATE
+# A call to "unset" with no arguments causes at least Solaris 10
+# /usr/xpg4/bin/sh and /bin/ksh to bail out.  So keep the unsets
+# deriving from the command substitution clustered with the other
+# ones.
+unset VISUAL EMAIL LANGUAGE COLUMNS $(perl -e '
+       my @env = keys %ENV;
+       my $ok = join("|", qw(
+               TRACE
+               DEBUG
+               USE_LOOKUP
+               TEST
+               .*_TEST
+               PROVE
+               VALGRIND
+               PERF_AGGREGATING_LATER
+       ));
+       my @vars = grep(/^GIT_/ && !/^GIT_($ok)/o, @env);
+       print join("\n", @vars);
+')
 GIT_AUTHOR_EMAIL=author@example.com
 GIT_AUTHOR_NAME='A U Thor'
-unset GIT_COMMITTER_DATE
 GIT_COMMITTER_EMAIL=committer@example.com
 GIT_COMMITTER_NAME='C O Mitter'
-unset GIT_DIFF_OPTS
-unset GIT_DIR
-unset GIT_EXTERNAL_DIFF
-unset GIT_INDEX_FILE
-unset GIT_OBJECT_DIRECTORY
-unset GIT_TRACE
-unset SHA1_FILE_DIRECTORIES
-unset SHA1_FILE_DIRECTORY
+GIT_MERGE_VERBOSITY=5
+GIT_MERGE_AUTOEDIT=no
+export GIT_MERGE_VERBOSITY GIT_MERGE_AUTOEDIT
 export GIT_AUTHOR_EMAIL GIT_AUTHOR_NAME
 export GIT_COMMITTER_EMAIL GIT_COMMITTER_NAME
-export EDITOR VISUAL
+export EDITOR
+
+# Protect ourselves from common misconfiguration to export
+# CDPATH into the environment
+unset CDPATH
+
+unset GREP_OPTIONS
+
+case $(echo $GIT_TRACE |tr "[A-Z]" "[a-z]") in
+       1|2|true)
+               echo "* warning: Some tests will not work if GIT_TRACE" \
+                       "is set as to trace on STDERR ! *"
+               echo "* warning: Please set GIT_TRACE to something" \
+                       "other than 1, 2 or true ! *"
+               ;;
+esac
+
+# Convenience
+#
+# A regexp to match 5 and 40 hexdigits
+_x05='[0-9a-f][0-9a-f][0-9a-f][0-9a-f][0-9a-f]'
+_x40="$_x05$_x05$_x05$_x05$_x05$_x05$_x05$_x05"
+
+# Zero SHA-1
+_z40=0000000000000000000000000000000000000000
+
+# Line feed
+LF='
+'
+
+export _x05 _x40 _z40 LF
 
 # Each test should start with something like this, after copyright notices:
 #
@@ -41,19 +108,15 @@ export EDITOR VISUAL
 # This test checks if command xyzzy does the right thing...
 # '
 # . ./test-lib.sh
-
-error () {
-       echo "* error: $*"
-       trap - exit
-       exit 1
-}
-
-say () {
-       echo "* $*"
-}
-
-test "${test_description}" != "" ||
-error "Test script did not set test_description."
+[ "x$ORIGINAL_TERM" != "xdumb" ] && (
+               TERM=$ORIGINAL_TERM &&
+               export TERM &&
+               [ -t 1 ] &&
+               tput bold >/dev/null 2>&1 &&
+               tput setaf 1 >/dev/null 2>&1 &&
+               tput sgr0 >/dev/null 2>&1
+       ) &&
+       color=t
 
 while test "$#" -ne 0
 do
@@ -62,19 +125,79 @@ do
                debug=t; shift ;;
        -i|--i|--im|--imm|--imme|--immed|--immedi|--immedia|--immediat|--immediate)
                immediate=t; shift ;;
+       -l|--l|--lo|--lon|--long|--long-|--long-t|--long-te|--long-tes|--long-test|--long-tests)
+               GIT_TEST_LONG=t; export GIT_TEST_LONG; shift ;;
        -h|--h|--he|--hel|--help)
-               echo "$test_description"
-               exit 0 ;;
+               help=t; shift ;;
        -v|--v|--ve|--ver|--verb|--verbo|--verbos|--verbose)
                verbose=t; shift ;;
-       --no-python)
-               no_python=t; shift ;;
+       -q|--q|--qu|--qui|--quie|--quiet)
+               # Ignore --quiet under a TAP::Harness. Saying how many tests
+               # passed without the ok/not ok details is always an error.
+               test -z "$HARNESS_ACTIVE" && quiet=t; shift ;;
+       --with-dashes)
+               with_dashes=t; shift ;;
+       --no-color)
+               color=; shift ;;
+       --va|--val|--valg|--valgr|--valgri|--valgrin|--valgrind)
+               valgrind=t; verbose=t; shift ;;
+       --tee)
+               shift ;; # was handled already
+       --root=*)
+               root=$(expr "z$1" : 'z[^=]*=\(.*\)')
+               shift ;;
        *)
-               break ;;
+               echo "error: unknown test option '$1'" >&2; exit 1 ;;
        esac
 done
 
+if test -n "$color"; then
+       say_color () {
+               (
+               TERM=$ORIGINAL_TERM
+               export TERM
+               case "$1" in
+                       error) tput bold; tput setaf 1;; # bold red
+                       skip)  tput bold; tput setaf 2;; # bold green
+                       pass)  tput setaf 2;;            # green
+                       info)  tput setaf 3;;            # brown
+                       *) test -n "$quiet" && return;;
+               esac
+               shift
+               printf "%s" "$*"
+               tput sgr0
+               echo
+               )
+       }
+else
+       say_color() {
+               test -z "$1" && test -n "$quiet" && return
+               shift
+               echo "$*"
+       }
+fi
+
+error () {
+       say_color error "error: $*"
+       GIT_EXIT_OK=t
+       exit 1
+}
+
+say () {
+       say_color info "$*"
+}
+
+test "${test_description}" != "" ||
+error "Test script did not set test_description."
+
+if test "$help" = "t"
+then
+       echo "$test_description"
+       exit 0
+fi
+
 exec 5>&1
+exec 6<&0
 if test "$verbose" = "t"
 then
        exec 4>&2 3>&1
@@ -84,138 +207,460 @@ fi
 
 test_failure=0
 test_count=0
+test_fixed=0
+test_broken=0
+test_success=0
 
-trap 'echo >&5 "FATAL: Unexpected exit with code $?"; exit 1' exit
+test_external_has_tap=0
 
+die () {
+       code=$?
+       if test -n "$GIT_EXIT_OK"
+       then
+               exit $code
+       else
+               echo >&5 "FATAL: Unexpected exit with code $code"
+               exit 1
+       fi
+}
+
+GIT_EXIT_OK=
+trap 'die' EXIT
+
+# The user-facing functions are loaded from a separate file so that
+# test_perf subshells can have them too
+. "${TEST_DIRECTORY:-.}"/test-lib-functions.sh
 
 # You are not expected to call test_ok_ and test_failure_ directly, use
 # the text_expect_* functions instead.
 
 test_ok_ () {
-       test_count=$(expr "$test_count" + 1)
-       say "  ok $test_count: $@"
+       test_success=$(($test_success + 1))
+       say_color "" "ok $test_count - $@"
 }
 
 test_failure_ () {
-       test_count=$(expr "$test_count" + 1)
-       test_failure=$(expr "$test_failure" + 1);
-       say "FAIL $test_count: $1"
+       test_failure=$(($test_failure + 1))
+       say_color error "not ok - $test_count $1"
        shift
-       echo "$@" | sed -e 's/^/        /'
-       test "$immediate" = "" || { trap - exit; exit 1; }
+       echo "$@" | sed -e 's/^/#       /'
+       test "$immediate" = "" || { GIT_EXIT_OK=t; exit 1; }
+}
+
+test_known_broken_ok_ () {
+       test_fixed=$(($test_fixed+1))
+       say_color "" "ok $test_count - $@ # TODO known breakage"
 }
 
+test_known_broken_failure_ () {
+       test_broken=$(($test_broken+1))
+       say_color skip "not ok $test_count - $@ # TODO known breakage"
+}
 
 test_debug () {
        test "$debug" = "" || eval "$1"
 }
 
-test_run_ () {
-       eval >&3 2>&4 "$1"
-       eval_ret="$?"
-       return 0
+test_eval_ () {
+       # This is a separate function because some tests use
+       # "return" to end a test_expect_success block early.
+       eval </dev/null >&3 2>&4 "$*"
 }
 
-test_expect_failure () {
-       test "$#" = 2 ||
-       error "bug in the test script: not 2 parameters to test-expect-failure"
-       say >&3 "expecting failure: $2"
-       test_run_ "$2"
-       if [ "$?" = 0 -a "$eval_ret" != 0 ]
+test_run_ () {
+       test_cleanup=:
+       expecting_failure=$2
+       test_eval_ "$1"
+       eval_ret=$?
+
+       if test -z "$immediate" || test $eval_ret = 0 || test -n "$expecting_failure"
        then
-               test_ok_ "$1"
-       else
-               test_failure_ "$@"
+               test_eval_ "$test_cleanup"
+       fi
+       if test "$verbose" = "t" && test -n "$HARNESS_ACTIVE"; then
+               echo ""
        fi
+       return "$eval_ret"
 }
 
-test_expect_success () {
-       test "$#" = 2 ||
-       error "bug in the test script: not 2 parameters to test-expect-success"
-       say >&3 "expecting success: $2"
-       test_run_ "$2"
-       if [ "$?" = 0 -a "$eval_ret" = 0 ]
+test_skip () {
+       test_count=$(($test_count+1))
+       to_skip=
+       for skp in $GIT_SKIP_TESTS
+       do
+               case $this_test.$test_count in
+               $skp)
+                       to_skip=t
+                       break
+               esac
+       done
+       if test -z "$to_skip" && test -n "$test_prereq" &&
+          ! test_have_prereq "$test_prereq"
        then
-               test_ok_ "$1"
-       else
-               test_failure_ "$@"
+               to_skip=t
        fi
+       case "$to_skip" in
+       t)
+               of_prereq=
+               if test "$missing_prereq" != "$test_prereq"
+               then
+                       of_prereq=" of $test_prereq"
+               fi
+
+               say_color skip >&3 "skipping test: $@"
+               say_color skip "ok $test_count # skip $1 (missing $missing_prereq${of_prereq})"
+               : true
+               ;;
+       *)
+               false
+               ;;
+       esac
 }
 
-test_expect_code () {
-       test "$#" = 3 ||
-       error "bug in the test script: not 3 parameters to test-expect-code"
-       say >&3 "expecting exit code $1: $3"
-       test_run_ "$3"
-       if [ "$?" = 0 -a "$eval_ret" = "$1" ]
-       then
-               test_ok_ "$2"
-       else
-               test_failure_ "$@"
-       fi
+# stub; perf-lib overrides it
+test_at_end_hook_ () {
+       :
 }
 
-# Most tests can use the created repository, but some amy need to create more.
-# Usage: test_create_repo <directory>
-test_create_repo () {
-       test "$#" = 1 ||
-       error "bug in the test script: not 1 parameter to test-create-repo"
-       owd=`pwd`
-       repo="$1"
-       mkdir "$repo"
-       cd "$repo" || error "Cannot setup test environment"
-       "$GIT_EXEC_PATH/git" init-db --template=$GIT_EXEC_PATH/templates/blt/ 2>/dev/null ||
-       error "cannot run git init-db -- have you built things yet?"
-       mv .git/hooks .git/hooks-disabled
-       cd "$owd"
-}
-       
 test_done () {
-       trap - exit
+       GIT_EXIT_OK=t
+
+       if test -z "$HARNESS_ACTIVE"; then
+               test_results_dir="$TEST_OUTPUT_DIRECTORY/test-results"
+               mkdir -p "$test_results_dir"
+               test_results_path="$test_results_dir/${0%.sh}-$$.counts"
+
+               cat >>"$test_results_path" <<-EOF
+               total $test_count
+               success $test_success
+               fixed $test_fixed
+               broken $test_broken
+               failed $test_failure
+
+               EOF
+       fi
+
+       if test "$test_fixed" != 0
+       then
+               say_color pass "# fixed $test_fixed known breakage(s)"
+       fi
+       if test "$test_broken" != 0
+       then
+               say_color error "# still have $test_broken known breakage(s)"
+               msg="remaining $(($test_count-$test_broken)) test(s)"
+       else
+               msg="$test_count test(s)"
+       fi
        case "$test_failure" in
        0)
-               # We could:
-               # cd .. && rm -fr trash
-               # but that means we forbid any tests that use their own
-               # subdirectory from calling test_done without coming back
-               # to where they started from.
-               # The Makefile provided will clean this test area so
-               # we will leave things as they are.
-
-               say "passed all $test_count test(s)"
+               # Maybe print SKIP message
+               [ -z "$skip_all" ] || skip_all=" # SKIP $skip_all"
+
+               if test $test_external_has_tap -eq 0; then
+                       say_color pass "# passed all $msg"
+                       say "1..$test_count$skip_all"
+               fi
+
+               test -d "$remove_trash" &&
+               cd "$(dirname "$remove_trash")" &&
+               rm -rf "$(basename "$remove_trash")"
+
+               test_at_end_hook_
+
                exit 0 ;;
 
        *)
-               say "failed $test_failure among $test_count test(s)"
+               if test $test_external_has_tap -eq 0; then
+                       say_color error "# failed $test_failure among $msg"
+                       say "1..$test_count"
+               fi
+
                exit 1 ;;
 
        esac
 }
 
 # Test the binaries we have just built.  The tests are kept in
-# t/ subdirectory and are run in trash subdirectory.
-PATH=$(pwd)/..:$PATH
-GIT_EXEC_PATH=$(pwd)/..
-export PATH GIT_EXEC_PATH
-
-# Similarly use ../compat/subprocess.py if our python does not
-# have subprocess.py on its own.
-PYTHON=`sed -e '1{
-       s/^#!//
-       q
-}' ../git-merge-recursive` || {
+# t/ subdirectory and are run in 'trash directory' subdirectory.
+if test -z "$TEST_DIRECTORY"
+then
+       # We allow tests to override this, in case they want to run tests
+       # outside of t/, e.g. for running tests on the test library
+       # itself.
+       TEST_DIRECTORY=$(pwd)
+fi
+if test -z "$TEST_OUTPUT_DIRECTORY"
+then
+       # Similarly, override this to store the test-results subdir
+       # elsewhere
+       TEST_OUTPUT_DIRECTORY=$TEST_DIRECTORY
+fi
+GIT_BUILD_DIR="$TEST_DIRECTORY"/..
+
+if test -n "$valgrind"
+then
+       make_symlink () {
+               test -h "$2" &&
+               test "$1" = "$(readlink "$2")" || {
+                       # be super paranoid
+                       if mkdir "$2".lock
+                       then
+                               rm -f "$2" &&
+                               ln -s "$1" "$2" &&
+                               rm -r "$2".lock
+                       else
+                               while test -d "$2".lock
+                               do
+                                       say "Waiting for lock on $2."
+                                       sleep 1
+                               done
+                       fi
+               }
+       }
+
+       make_valgrind_symlink () {
+               # handle only executables, unless they are shell libraries that
+               # need to be in the exec-path.  We will just use "#!" as a
+               # guess for a shell-script, since we have no idea what the user
+               # may have configured as the shell path.
+               test -x "$1" ||
+               test "#!" = "$(head -c 2 <"$1")" ||
+               return;
+
+               base=$(basename "$1")
+               symlink_target=$GIT_BUILD_DIR/$base
+               # do not override scripts
+               if test -x "$symlink_target" &&
+                   test ! -d "$symlink_target" &&
+                   test "#!" != "$(head -c 2 < "$symlink_target")"
+               then
+                       symlink_target=../valgrind.sh
+               fi
+               case "$base" in
+               *.sh|*.perl)
+                       symlink_target=../unprocessed-script
+               esac
+               # create the link, or replace it if it is out of date
+               make_symlink "$symlink_target" "$GIT_VALGRIND/bin/$base" || exit
+       }
+
+       # override all git executables in TEST_DIRECTORY/..
+       GIT_VALGRIND=$TEST_DIRECTORY/valgrind
+       mkdir -p "$GIT_VALGRIND"/bin
+       for file in $GIT_BUILD_DIR/git* $GIT_BUILD_DIR/test-*
+       do
+               make_valgrind_symlink $file
+       done
+       # special-case the mergetools loadables
+       make_symlink "$GIT_BUILD_DIR"/mergetools "$GIT_VALGRIND/bin/mergetools"
+       OLDIFS=$IFS
+       IFS=:
+       for path in $PATH
+       do
+               ls "$path"/git-* 2> /dev/null |
+               while read file
+               do
+                       make_valgrind_symlink "$file"
+               done
+       done
+       IFS=$OLDIFS
+       PATH=$GIT_VALGRIND/bin:$PATH
+       GIT_EXEC_PATH=$GIT_VALGRIND/bin
+       export GIT_VALGRIND
+elif test -n "$GIT_TEST_INSTALLED" ; then
+       GIT_EXEC_PATH=$($GIT_TEST_INSTALLED/git --exec-path)  ||
+       error "Cannot run git from $GIT_TEST_INSTALLED."
+       PATH=$GIT_TEST_INSTALLED:$GIT_BUILD_DIR:$PATH
+       GIT_EXEC_PATH=${GIT_TEST_EXEC_PATH:-$GIT_EXEC_PATH}
+else # normal case, use ../bin-wrappers only unless $with_dashes:
+       git_bin_dir="$GIT_BUILD_DIR/bin-wrappers"
+       if ! test -x "$git_bin_dir/git" ; then
+               if test -z "$with_dashes" ; then
+                       say "$git_bin_dir/git is not executable; using GIT_EXEC_PATH"
+               fi
+               with_dashes=t
+       fi
+       PATH="$git_bin_dir:$PATH"
+       GIT_EXEC_PATH=$GIT_BUILD_DIR
+       if test -n "$with_dashes" ; then
+               PATH="$GIT_BUILD_DIR:$PATH"
+       fi
+fi
+GIT_TEMPLATE_DIR="$GIT_BUILD_DIR"/templates/blt
+unset GIT_CONFIG
+GIT_CONFIG_NOSYSTEM=1
+GIT_ATTR_NOSYSTEM=1
+export PATH GIT_EXEC_PATH GIT_TEMPLATE_DIR GIT_CONFIG_NOSYSTEM GIT_ATTR_NOSYSTEM
+
+. "$GIT_BUILD_DIR"/GIT-BUILD-OPTIONS
+
+if test -z "$GIT_TEST_CMP"
+then
+       if test -n "$GIT_TEST_CMP_USE_COPIED_CONTEXT"
+       then
+               GIT_TEST_CMP="$DIFF -c"
+       else
+               GIT_TEST_CMP="$DIFF -u"
+       fi
+fi
+
+GITPERLLIB="$GIT_BUILD_DIR"/perl/blib/lib:"$GIT_BUILD_DIR"/perl/blib/arch/auto/Git
+export GITPERLLIB
+test -d "$GIT_BUILD_DIR"/templates/blt || {
        error "You haven't built things yet, have you?"
 }
-"$PYTHON" -c 'import subprocess' 2>/dev/null || {
-       PYTHONPATH=$(pwd)/../compat
-       export PYTHONPATH
+
+if test -z "$GIT_TEST_INSTALLED" && test -z "$NO_PYTHON"
+then
+       GITPYTHONLIB="$GIT_BUILD_DIR/git_remote_helpers/build/lib"
+       export GITPYTHONLIB
+       test -d "$GIT_BUILD_DIR"/git_remote_helpers/build || {
+               error "You haven't built git_remote_helpers yet, have you?"
+       }
+fi
+
+if ! test -x "$GIT_BUILD_DIR"/test-chmtime; then
+       echo >&2 'You need to build test-chmtime:'
+       echo >&2 'Run "make test-chmtime" in the source (toplevel) directory'
+       exit 1
+fi
+
+# Test repository
+test="trash directory.$(basename "$0" .sh)"
+test -n "$root" && test="$root/$test"
+case "$test" in
+/*) TRASH_DIRECTORY="$test" ;;
+ *) TRASH_DIRECTORY="$TEST_OUTPUT_DIRECTORY/$test" ;;
+esac
+test ! -z "$debug" || remove_trash=$TRASH_DIRECTORY
+rm -fr "$test" || {
+       GIT_EXIT_OK=t
+       echo >&5 "FATAL: Cannot prepare test area"
+       exit 1
 }
-test -d ../templates/blt || {
-       error "You haven't built things yet, have you?"
+
+HOME="$TRASH_DIRECTORY"
+export HOME
+
+if test -z "$TEST_NO_CREATE_REPO"; then
+       test_create_repo "$test"
+else
+       mkdir -p "$test"
+fi
+# Use -P to resolve symlinks in our working directory so that the cwd
+# in subprocesses like git equals our $PWD (for pathname comparisons).
+cd -P "$test" || exit 1
+
+this_test=${0##*/}
+this_test=${this_test%%-*}
+for skp in $GIT_SKIP_TESTS
+do
+       case "$this_test" in
+       $skp)
+               say_color skip >&3 "skipping test $this_test altogether"
+               skip_all="skip all tests in $this_test"
+               test_done
+       esac
+done
+
+# Provide an implementation of the 'yes' utility
+yes () {
+       if test $# = 0
+       then
+               y=y
+       else
+               y="$*"
+       fi
+
+       while echo "$y"
+       do
+               :
+       done
 }
 
-# Test repository
-test=trash
-rm -fr "$test"
-test_create_repo $test
-cd "$test"
+# Fix some commands on Windows
+case $(uname -s) in
+*MINGW*)
+       # Windows has its own (incompatible) sort and find
+       sort () {
+               /usr/bin/sort "$@"
+       }
+       find () {
+               /usr/bin/find "$@"
+       }
+       sum () {
+               md5sum "$@"
+       }
+       # git sees Windows-style pwd
+       pwd () {
+               builtin pwd -W
+       }
+       # no POSIX permissions
+       # backslashes in pathspec are converted to '/'
+       # exec does not inherit the PID
+       test_set_prereq MINGW
+       test_set_prereq SED_STRIPS_CR
+       ;;
+*CYGWIN*)
+       test_set_prereq POSIXPERM
+       test_set_prereq EXECKEEPSPID
+       test_set_prereq NOT_MINGW
+       test_set_prereq SED_STRIPS_CR
+       ;;
+*)
+       test_set_prereq POSIXPERM
+       test_set_prereq BSLASHPSPEC
+       test_set_prereq EXECKEEPSPID
+       test_set_prereq NOT_MINGW
+       ;;
+esac
+
+( COLUMNS=1 && test $COLUMNS = 1 ) && test_set_prereq COLUMNS_CAN_BE_1
+test -z "$NO_PERL" && test_set_prereq PERL
+test -z "$NO_PYTHON" && test_set_prereq PYTHON
+test -n "$USE_LIBPCRE" && test_set_prereq LIBPCRE
+test -z "$NO_GETTEXT" && test_set_prereq GETTEXT
+
+# Can we rely on git's output in the C locale?
+if test -n "$GETTEXT_POISON"
+then
+       GIT_GETTEXT_POISON=YesPlease
+       export GIT_GETTEXT_POISON
+       test_set_prereq GETTEXT_POISON
+else
+       test_set_prereq C_LOCALE_OUTPUT
+fi
+
+# Use this instead of test_cmp to compare files that contain expected and
+# actual output from git commands that can be translated.  When running
+# under GETTEXT_POISON this pretends that the command produced expected
+# results.
+test_i18ncmp () {
+       test -n "$GETTEXT_POISON" || test_cmp "$@"
+}
+
+# Use this instead of "grep expected-string actual" to see if the
+# output from a git command that can be translated either contains an
+# expected string, or does not contain an unwanted one.  When running
+# under GETTEXT_POISON this pretends that the command produced expected
+# results.
+test_i18ngrep () {
+       if test -n "$GETTEXT_POISON"
+       then
+           : # pretend success
+       elif test "x!" = "x$1"
+       then
+               shift
+               ! grep "$@"
+       else
+               grep "$@"
+       fi
+}
+
+# test whether the filesystem supports symbolic links
+ln -s x y 2>/dev/null && test -h y 2>/dev/null && test_set_prereq SYMLINKS
+rm -f y
+
+# When the tests are run as root, permission tests will report that
+# things are writable when they shouldn't be.
+test -w / || test_set_prereq SANITY