Merge branch 'gc/http-with-non-ascii-username-url'
[gitweb.git] / t / test-lib.sh
index 29fd7209cf1aff851f0f1f95ccecd0d949b3da6c..48fa5160045d0bb2c14d034422b2ee5170889ec5 100644 (file)
@@ -238,14 +238,51 @@ test_set_editor () {
 }
 
 test_decode_color () {
-       sed     -e 's/.\[1m/<WHITE>/g' \
-               -e 's/.\[31m/<RED>/g' \
-               -e 's/.\[32m/<GREEN>/g' \
-               -e 's/.\[33m/<YELLOW>/g' \
-               -e 's/.\[34m/<BLUE>/g' \
-               -e 's/.\[35m/<MAGENTA>/g' \
-               -e 's/.\[36m/<CYAN>/g' \
-               -e 's/.\[m/<RESET>/g'
+       awk '
+               function name(n) {
+                       if (n == 0) return "RESET";
+                       if (n == 1) return "BOLD";
+                       if (n == 30) return "BLACK";
+                       if (n == 31) return "RED";
+                       if (n == 32) return "GREEN";
+                       if (n == 33) return "YELLOW";
+                       if (n == 34) return "BLUE";
+                       if (n == 35) return "MAGENTA";
+                       if (n == 36) return "CYAN";
+                       if (n == 37) return "WHITE";
+                       if (n == 40) return "BLACK";
+                       if (n == 41) return "BRED";
+                       if (n == 42) return "BGREEN";
+                       if (n == 43) return "BYELLOW";
+                       if (n == 44) return "BBLUE";
+                       if (n == 45) return "BMAGENTA";
+                       if (n == 46) return "BCYAN";
+                       if (n == 47) return "BWHITE";
+               }
+               {
+                       while (match($0, /\x1b\[[0-9;]*m/) != 0) {
+                               printf "%s<", substr($0, 1, RSTART-1);
+                               codes = substr($0, RSTART+2, RLENGTH-3);
+                               if (length(codes) == 0)
+                                       printf "%s", name(0)
+                               else {
+                                       n = split(codes, ary, ";");
+                                       sep = "";
+                                       for (i = 1; i <= n; i++) {
+                                               printf "%s%s", sep, name(ary[i]);
+                                               sep = ";"
+                                       }
+                               }
+                               printf ">";
+                               $0 = substr($0, RSTART + RLENGTH, length($0) - RSTART - RLENGTH + 1);
+                       }
+                       print
+               }
+       '
+}
+
+nul_to_q () {
+       perl -pe 'y/\000/Q/'
 }
 
 q_to_nul () {
@@ -268,6 +305,17 @@ remove_cr () {
        tr '\015' Q | sed -e 's/Q$//'
 }
 
+# In some bourne shell implementations, the "unset" builtin returns
+# nonzero status when a variable to be unset was not set in the first
+# place.
+#
+# Use sane_unset when that should not be considered an error.
+
+sane_unset () {
+       unset "$@"
+       return 0
+}
+
 test_tick () {
        if test -z "${test_tick+set}"
        then
@@ -331,12 +379,44 @@ test_set_prereq () {
 satisfied=" "
 
 test_have_prereq () {
-       case $satisfied in
-       *" $1 "*)
-               : yes, have it ;;
-       *)
-               ! : nope ;;
+       # prerequisites can be concatenated with ','
+       save_IFS=$IFS
+       IFS=,
+       set -- $*
+       IFS=$save_IFS
+
+       total_prereq=0
+       ok_prereq=0
+       missing_prereq=
+
+       for prerequisite
+       do
+               total_prereq=$(($total_prereq + 1))
+               case $satisfied in
+               *" $prerequisite "*)
+                       ok_prereq=$(($ok_prereq + 1))
+                       ;;
+               *)
+                       # Keep a list of missing prerequisites
+                       if test -z "$missing_prereq"
+                       then
+                               missing_prereq=$prerequisite
+                       else
+                               missing_prereq="$prerequisite,$missing_prereq"
+                       fi
+               esac
+       done
+
+       test $total_prereq = $ok_prereq
+}
+
+test_declared_prereq () {
+       case ",$test_prereq," in
+       *,$1,*)
+               return 0
+               ;;
        esac
+       return 1
 }
 
 # You are not expected to call test_ok_ and test_failure_ directly, use
@@ -391,15 +471,21 @@ test_skip () {
                        break
                esac
        done
-       if test -z "$to_skip" && test -n "$prereq" &&
-          ! test_have_prereq "$prereq"
+       if test -z "$to_skip" && test -n "$test_prereq" &&
+          ! test_have_prereq "$test_prereq"
        then
                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"
+               say_color skip "ok $test_count # skip $1 (missing $missing_prereq${of_prereq})"
                : true
                ;;
        *)
@@ -409,9 +495,10 @@ test_skip () {
 }
 
 test_expect_failure () {
-       test "$#" = 3 && { prereq=$1; shift; } || prereq=
+       test "$#" = 3 && { test_prereq=$1; shift; } || test_prereq=
        test "$#" = 2 ||
        error "bug in the test script: not 2 or 3 parameters to test-expect-failure"
+       export test_prereq
        if ! test_skip "$@"
        then
                say >&3 "checking known breakage: $2"
@@ -427,9 +514,10 @@ test_expect_failure () {
 }
 
 test_expect_success () {
-       test "$#" = 3 && { prereq=$1; shift; } || prereq=
+       test "$#" = 3 && { test_prereq=$1; shift; } || test_prereq=
        test "$#" = 2 ||
        error "bug in the test script: not 2 or 3 parameters to test-expect-success"
+       export test_prereq
        if ! test_skip "$@"
        then
                say >&3 "expecting success: $2"
@@ -444,24 +532,6 @@ test_expect_success () {
        echo >&3 ""
 }
 
-test_expect_code () {
-       test "$#" = 4 && { prereq=$1; shift; } || prereq=
-       test "$#" = 3 ||
-       error "bug in the test script: not 3 or 4 parameters to test-expect-code"
-       if ! test_skip "$@"
-       then
-               say >&3 "expecting exit code $1: $3"
-               test_run_ "$3"
-               if [ "$?" = 0 -a "$eval_ret" = "$1" ]
-               then
-                       test_ok_ "$2"
-               else
-                       test_failure_ "$@"
-               fi
-       fi
-       echo >&3 ""
-}
-
 # test_external runs external test scripts that provide continuous
 # test output about their progress, and succeeds/fails on
 # zero/non-zero exit code.  It outputs the test output on stdout even
@@ -471,11 +541,12 @@ test_expect_code () {
 # Usage: test_external description command arguments...
 # Example: test_external 'Perl API' perl ../path/to/test.pl
 test_external () {
-       test "$#" = 4 && { prereq=$1; shift; } || prereq=
+       test "$#" = 4 && { test_prereq=$1; shift; } || test_prereq=
        test "$#" = 3 ||
        error >&5 "bug in the test script: not 3 or 4 parameters to test_external"
        descr="$1"
        shift
+       export test_prereq
        if ! test_skip "$descr" "$@"
        then
                # Announce the script to reduce confusion about the
@@ -545,6 +616,60 @@ test_external_without_stderr () {
        fi
 }
 
+# debugging-friendly alternatives to "test [-f|-d|-e]"
+# The commands test the existence or non-existence of $1. $2 can be
+# given to provide a more precise diagnosis.
+test_path_is_file () {
+       if ! [ -f "$1" ]
+       then
+               echo "File $1 doesn't exist. $*"
+               false
+       fi
+}
+
+test_path_is_dir () {
+       if ! [ -d "$1" ]
+       then
+               echo "Directory $1 doesn't exist. $*"
+               false
+       fi
+}
+
+test_path_is_missing () {
+       if [ -e "$1" ]
+       then
+               echo "Path exists:"
+               ls -ld "$1"
+               if [ $# -ge 1 ]; then
+                       echo "$*"
+               fi
+               false
+       fi
+}
+
+# test_line_count checks that a file has the number of lines it
+# ought to. For example:
+#
+#      test_expect_success 'produce exactly one line of output' '
+#              do something >output &&
+#              test_line_count = 1 output
+#      '
+#
+# is like "test $(wc -l <output) = 1" except that it passes the
+# output through when the number of lines is wrong.
+
+test_line_count () {
+       if test $# != 3
+       then
+               error "bug in the test script: not 3 parameters to test_line_count"
+       elif ! test $(wc -l <"$3") "$1" "$2"
+       then
+               echo "test_line_count: line count for $3 !$1 $2"
+               cat "$3"
+               return 1
+       fi
+}
+
 # This is not among top-level (test_expect_success | test_expect_failure)
 # but is a prefix that can be used in the test script, like:
 #
@@ -559,7 +684,18 @@ test_external_without_stderr () {
 
 test_must_fail () {
        "$@"
-       test $? -gt 0 -a $? -le 129 -o $? -gt 192
+       exit_code=$?
+       if test $exit_code = 0; then
+               echo >&2 "test_must_fail: command succeeded: $*"
+               return 1
+       elif test $exit_code -gt 129 -a $exit_code -le 192; then
+               echo >&2 "test_must_fail: died by signal: $*"
+               return 1
+       elif test $exit_code = 127; then
+               echo >&2 "test_must_fail: command not found: $*"
+               return 1
+       fi
+       return 0
 }
 
 # Similar to test_must_fail, but tolerates success, too.  This is
@@ -575,7 +711,37 @@ test_must_fail () {
 
 test_might_fail () {
        "$@"
-       test $? -ge 0 -a $? -le 129 -o $? -gt 192
+       exit_code=$?
+       if test $exit_code -gt 129 -a $exit_code -le 192; then
+               echo >&2 "test_might_fail: died by signal: $*"
+               return 1
+       elif test $exit_code = 127; then
+               echo >&2 "test_might_fail: command not found: $*"
+               return 1
+       fi
+       return 0
+}
+
+# Similar to test_must_fail and test_might_fail, but check that a
+# given command exited with a given exit code. Meant to be used as:
+#
+#      test_expect_success 'Merge with d/f conflicts' '
+#              test_expect_code 1 git merge "merge msg" B master
+#      '
+
+test_expect_code () {
+       want_code=$1
+       shift
+       "$@"
+       exit_code=$?
+       if test $exit_code = $want_code
+       then
+               echo >&2 "test_expect_code: command exited with $exit_code: $*"
+               return 0
+       else
+               echo >&2 "test_expect_code: command exited with $exit_code, we wanted $want_code $*"
+               return 1
+       fi
 }
 
 # test_cmp is a helper function to compare actual and expected output.
@@ -625,28 +791,31 @@ test_when_finished () {
 test_create_repo () {
        test "$#" = 1 ||
        error "bug in the test script: not 1 parameter to test-create-repo"
-       owd=`pwd`
        repo="$1"
        mkdir -p "$repo"
-       cd "$repo" || error "Cannot setup test environment"
-       "$GIT_EXEC_PATH/git-init" "--template=$TEST_DIRECTORY/../templates/blt/" >&3 2>&4 ||
-       error "cannot run git init -- have you built things yet?"
-       mv .git/hooks .git/hooks-disabled
-       cd "$owd"
+       (
+               cd "$repo" || error "Cannot setup test environment"
+               "$GIT_EXEC_PATH/git-init" "--template=$GIT_BUILD_DIR/templates/blt/" >&3 2>&4 ||
+               error "cannot run git init -- have you built things yet?"
+               mv .git/hooks .git/hooks-disabled
+       ) || exit
 }
 
 test_done () {
        GIT_EXIT_OK=t
-       test_results_dir="$TEST_DIRECTORY/test-results"
-       mkdir -p "$test_results_dir"
-       test_results_path="$test_results_dir/${0%.sh}-$$.counts"
 
-       echo "total $test_count" >> $test_results_path
-       echo "success $test_success" >> $test_results_path
-       echo "fixed $test_fixed" >> $test_results_path
-       echo "broken $test_broken" >> $test_results_path
-       echo "failed $test_failure" >> $test_results_path
-       echo "" >> $test_results_path
+       if test -z "$HARNESS_ACTIVE"; then
+               test_results_dir="$TEST_DIRECTORY/test-results"
+               mkdir -p "$test_results_dir"
+               test_results_path="$test_results_dir/${0%.sh}-$$.counts"
+
+               echo "total $test_count" >> $test_results_path
+               echo "success $test_success" >> $test_results_path
+               echo "fixed $test_fixed" >> $test_results_path
+               echo "broken $test_broken" >> $test_results_path
+               echo "failed $test_failure" >> $test_results_path
+               echo "" >> $test_results_path
+       fi
 
        if test "$test_fixed" != 0
        then
@@ -688,7 +857,15 @@ test_done () {
 
 # Test the binaries we have just built.  The tests are kept in
 # t/ subdirectory and are run in 'trash directory' subdirectory.
-TEST_DIRECTORY=$(pwd)
+if test -z "$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=$(pwd)
+fi
+GIT_BUILD_DIR="$TEST_DIRECTORY"/..
+
 if test -n "$valgrind"
 then
        make_symlink () {
@@ -715,7 +892,7 @@ then
                test -x "$1" || return
 
                base=$(basename "$1")
-               symlink_target=$TEST_DIRECTORY/../$base
+               symlink_target=$GIT_BUILD_DIR/$base
                # do not override scripts
                if test -x "$symlink_target" &&
                    test ! -d "$symlink_target" &&
@@ -734,7 +911,7 @@ then
        # override all git executables in TEST_DIRECTORY/..
        GIT_VALGRIND=$TEST_DIRECTORY/valgrind
        mkdir -p "$GIT_VALGRIND"/bin
-       for file in $TEST_DIRECTORY/../git* $TEST_DIRECTORY/../test-*
+       for file in $GIT_BUILD_DIR/git* $GIT_BUILD_DIR/test-*
        do
                make_valgrind_symlink $file
        done
@@ -755,10 +932,10 @@ then
 elif test -n "$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:$TEST_DIRECTORY/..:$PATH
+       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 $with_dashes:
-       git_bin_dir="$TEST_DIRECTORY/../bin-wrappers"
+       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"
@@ -766,18 +943,18 @@ else # normal case, use ../bin-wrappers only unless $with_dashes:
                with_dashes=t
        fi
        PATH="$git_bin_dir:$PATH"
-       GIT_EXEC_PATH=$TEST_DIRECTORY/..
+       GIT_EXEC_PATH=$GIT_BUILD_DIR
        if test -n "$with_dashes" ; then
-               PATH="$TEST_DIRECTORY/..:$PATH"
+               PATH="$GIT_BUILD_DIR:$PATH"
        fi
 fi
-GIT_TEMPLATE_DIR=$(pwd)/../templates/blt
+GIT_TEMPLATE_DIR="$GIT_BUILD_DIR"/templates/blt
 unset GIT_CONFIG
 GIT_CONFIG_NOSYSTEM=1
 GIT_CONFIG_NOGLOBAL=1
 export PATH GIT_EXEC_PATH GIT_TEMPLATE_DIR GIT_CONFIG_NOSYSTEM GIT_CONFIG_NOGLOBAL
 
-. ../GIT-BUILD-OPTIONS
+. "$GIT_BUILD_DIR"/GIT-BUILD-OPTIONS
 
 if test -z "$GIT_TEST_CMP"
 then
@@ -789,22 +966,22 @@ then
        fi
 fi
 
-GITPERLLIB=$(pwd)/../perl/blib/lib:$(pwd)/../perl/blib/arch/auto/Git
+GITPERLLIB="$GIT_BUILD_DIR"/perl/blib/lib:"$GIT_BUILD_DIR"/perl/blib/arch/auto/Git
 export GITPERLLIB
-test -d ../templates/blt || {
+test -d "$GIT_BUILD_DIR"/templates/blt || {
        error "You haven't built things yet, have you?"
 }
 
 if test -z "$GIT_TEST_INSTALLED" && test -z "$NO_PYTHON"
 then
-       GITPYTHONLIB="$(pwd)/../git_remote_helpers/build/lib"
+       GITPYTHONLIB="$GIT_BUILD_DIR/git_remote_helpers/build/lib"
        export GITPYTHONLIB
-       test -d ../git_remote_helpers/build || {
+       test -d "$GIT_BUILD_DIR"/git_remote_helpers/build || {
                error "You haven't built git_remote_helpers yet, have you?"
        }
 fi
 
-if ! test -x ../test-chmtime; then
+if ! test -x "$GIT_BUILD_DIR"/test-chmtime; then
        echo >&2 'You need to build test-chmtime:'
        echo >&2 'Run "make test-chmtime" in the source (toplevel) directory'
        exit 1
@@ -829,6 +1006,9 @@ test_create_repo "$test"
 # in subprocesses like git equals our $PWD (for pathname comparisons).
 cd -P "$test" || exit 1
 
+HOME=$(pwd)
+export HOME
+
 this_test=${0##*/}
 this_test=${this_test%%-*}
 for skp in $GIT_SKIP_TESTS
@@ -876,11 +1056,13 @@ case $(uname -s) in
        # no POSIX permissions
        # backslashes in pathspec are converted to '/'
        # exec does not inherit the PID
+       test_set_prereq MINGW
        ;;
 *)
        test_set_prereq POSIXPERM
        test_set_prereq BSLASHPSPEC
        test_set_prereq EXECKEEPSPID
+       test_set_prereq NOT_MINGW
        ;;
 esac
 
@@ -890,3 +1072,7 @@ test -z "$NO_PYTHON" && test_set_prereq PYTHON
 # 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