diff --git a/xml/regression.sh b/xml/regression.sh index f0e63e1434..5add49f822 100755 --- a/xml/regression.sh +++ b/xml/regression.sh @@ -1,677 +1,703 @@ #!/bin/sh # # Copyright 2018-2024 the Pacemaker project contributors # # The version control history for this file may have further details. # # This source code is licensed under the GNU General Public License version 2 # or later (GPLv2+) WITHOUT ANY WARRANTY. set -eu if [ ! -d "test-2" ]; then echo "$0 must be run from the xml subdirectory of a source tree" exit 1 fi DIFF="diff -u" DIFF_PAGER="less -LRX" RNG_VALIDATOR="xmllint --noout --relaxng" XSLT_PROCESSOR="xsltproc --nonet" # Default tests tests="test2to3 test2to3enter test2to3leave test2to3roundtrip cts_scheduler" # # commons # emit_result() { _er_howmany=${1:?} # how many errors (0/anything else incl. strings) _er_subject=${2:?} _er_prefix=${3-} if [ -n "$_er_prefix" ]; then _er_prefix="${_er_prefix}: " fi if [ "$_er_howmany" = "0" ]; then printf "%s%s finished OK\n" "${_er_prefix}" "${_er_subject}" else printf "%s%s encountered ${_er_howmany} errors\n" \ "${_er_prefix}" "${_er_subject}" fi } emit_error() { _ee_msg=${1:?} printf "%s\n" "${_ee_msg}" >&2 } # returns 1 + floor of base 2 logaritm for _lo0r_i in 1...255, # or 0 for _lo0r_i = 0 log2_or_0_return() { _lo0r_i=${1:?} return $(((!(_lo0r_i >> 1) && _lo0r_i) * 1 \ + (!(_lo0r_i >> 2) && _lo0r_i & (1 << 1)) * 2 \ + (!(_lo0r_i >> 3) && _lo0r_i & (1 << 2)) * 3 \ + (!(_lo0r_i >> 4) && _lo0r_i & (1 << 3)) * 4 \ + (!(_lo0r_i >> 5) && _lo0r_i & (1 << 4)) * 5 \ + (!(_lo0r_i >> 6) && _lo0r_i & (1 << 5)) * 6 \ + (!(_lo0r_i >> 7) && _lo0r_i & (1 << 6)) * 7 \ + !!(_lo0r_i >> 7) * 7 )) } # rough addition of two base 2 logarithms log2_or_0_add() { _lo0a_op1=${1:?} _lo0a_op2=${2:?} if [ "$_lo0a_op1" -gt "$_lo0a_op2" ]; then return ${_lo0a_op1} elif [ "$_lo0a_op2" -gt "$_lo0a_op1" ]; then return ${_lo0a_op2} elif [ "$_lo0a_op1" -gt 0 ]; then return $((_lo0a_op1 + 1)) else return ${_lo0a_op1} fi } # # test phases # # -r ... whether to remove referential files as well # stdin: input file per line test_cleaner() { _tc_cleanref=0 while [ $# -gt 0 ]; do case "$1" in -r) _tc_cleanref=1;; esac shift done while read _tc_origin; do _tc_origin=${_tc_origin%.*} rm -f "${_tc_origin}.up" "${_tc_origin}.up.err" rm -f "$(dirname "${_tc_origin}")/.$(basename "${_tc_origin}").up" if [ "$_tc_cleanref" -eq 1 ]; then rm -f "${_tc_origin}.ref" "${_tc_origin}.ref.err" fi done } test_explanation() { _tsc_template= while [ $# -gt 0 ]; do case "$1" in -o=*) _tsc_template="upgrade-${1#-o=}.xsl";; esac shift done ${XSLT_PROCESSOR} upgrade-detail.xsl "${_tsc_template}" } +cleanup_module_error() { + # Work around a libxml2 bug. At least as of libxslt-1.1.41 and + # libxml2-2.10.4, if the stylesheet contains a user-defined top-level + # element (that is, one with a namespace other than the XSL namespace), + # libxslt tries to load the namespace URI as an XML module. If this fails, + # libxml2 logs a "module error: failed to open ..." message. + # + # This appears to be fixed in libxml2 v2.13 with commit ecb4c9fb. + sed "/module error/d" "$1" > "$1.new" + mv -- "$1.new" "$1" +} + # stdout: filename of the transformed file test_runner_upgrade() { _tru_template=${1:?} _tru_source=${2:?} # filename _tru_mode=${3:?} # extra modes wrt. "referential" outcome, see below _tru_ref="${_tru_source%.*}.ref" if [ "$((_tru_mode & (1 << 0)))" -ne 0 ] || [ -f "${_tru_ref}.err" ]; then _tru_ref_err="${_tru_ref}.err" else _tru_ref_err=/dev/null fi + _tru_proc_rc=0 _tru_diff_rc=0 _tru_target="${_tru_source%.*}.up" _tru_target_err="${_tru_target}.err" if [ "$((_tru_mode & (1 << 2)))" -eq 0 ]; then ${XSLT_PROCESSOR} "${_tru_template}" "${_tru_source}" \ > "${_tru_target}" 2> "${_tru_target_err}" \ - || { _tru_ref=$?; echo "${_tru_target_err}" - return ${_tru_ref}; } + || _tru_proc_rc=$? + + cleanup_module_error "$_tru_target_err" + + if [ "$_tru_proc_rc" -ne 0 ]; then + echo "$_tru_target_err" + return "$_tru_proc_rc" + fi else # when -B (deblanked outcomes handling) requested, we: # - drop blanks from the source XML # (effectively emulating pacemaker handling) # - re-drop blanks from the XSLT outcome, # which is compared with referential outcome # processed with even greedier custom deblanking # (extraneous inter-element whitespace like blank # lines will not get removed otherwise, see lower) xmllint --noblanks "${_tru_source}" \ | ${XSLT_PROCESSOR} "${_tru_template}" - \ > "${_tru_target}" 2> "${_tru_target_err}" \ - || { _tru_ref=$?; echo "${_tru_target_err}" - return ${_tru_ref}; } + || _tru_proc_rc=$? + + cleanup_module_error "$_tru_target_err" + + if [ "$_tru_proc_rc" -ne 0 ]; then + echo "$_tru_target_err" + return "$_tru_proc_rc" + fi + # reusing variable no longer needed _tru_template="$(dirname "${_tru_target}")" _tru_template="${_tru_template}/.$(basename "${_tru_target}")" mv "${_tru_target}" "${_tru_template}" ${XSLT_PROCESSOR} - "${_tru_template}" > "${_tru_target}" < EOF fi # only respond with the flags except for "-B", i.e., when both: # - _tru_mode non-zero # - "-B" in _tru_mode is zero (hence non-zero when flipped with XOR) if [ "$((_tru_mode * ((_tru_mode ^ (1 << 2)) & (1 << 2))))" -ne 0 ]; then if [ "$((_tru_mode & (1 << 0)))" -ne 0 ]; then cp -a "${_tru_target}" "${_tru_ref}" cp -a "${_tru_target_err}" "${_tru_ref_err}" fi if [ "$((_tru_mode & (1 << 1)))" -ne 0 ]; then { ${DIFF} "${_tru_source}" "${_tru_ref}" \ && printf '\n(files match)\n'; } | ${DIFF_PAGER} >&2 if [ $? -ne 0 ]; then printf "\npager failure\n" >&2 return 1 fi printf '\nIs comparison OK? ' >&2 if read _tru_answer &2; return 1;; esac else return 1 fi fi elif [ -f "$_tru_ref" ] && [ -e "$_tru_ref_err" ]; then if [ "$((_tru_mode & (1 << 2)))" -eq 0 ]; then _output=$(cat "$_tru_ref") else _output=$($XSLT_PROCESSOR - "$_tru_ref" < EOF ) fi echo "$_output" | $DIFF - "$_tru_target" >&2 || _tru_diff_rc=$? if [ "$_tru_diff_rc" -eq 0 ]; then $DIFF "$_tru_ref_err" "$_tru_target_err" >&2 || _tru_diff_rc=$? fi if [ "$_tru_diff_rc" -ne 0 ]; then emit_error "Outputs differ from referential ones" echo "/dev/null" return 1 fi else emit_error "Referential file(s) missing: ${_tru_ref}" echo "/dev/null" return 1 fi echo "${_tru_target}" } test_runner_validate() { _trv_schema=${1:?} _trv_target=${2:?} # filename if ! ${RNG_VALIDATOR} "${_trv_schema}" "${_trv_target}" \ 2>/dev/null; then ${RNG_VALIDATOR} "${_trv_schema}" "${_trv_target}" fi } # -a= ... action modifier completing template name (e.g. 2.10-(enter|leave)) # -o= ... which conventional version to deem as the transform origin # -t= ... which conventional version to deem as the transform target # -B # -D # -G ... see usage # stdin: input file per line test_runner() { _tr_mode=0 _tr_ret=0 _tr_action= _tr_schema_o= _tr_schema_t= _tr_target= _tr_template= while [ $# -gt 0 ]; do case "$1" in -a=*) _tr_action="${1#-a=}";; -o=*) _tr_template="${1#-o=}" _tr_schema_o="pacemaker-${1#-o=}.rng";; -t=*) _tr_schema_t="pacemaker-${1#-t=}.rng";; -G) _tr_mode=$((_tr_mode | (1 << 0)));; -D) _tr_mode=$((_tr_mode | (1 << 1)));; -B) _tr_mode=$((_tr_mode | (1 << 2)));; esac shift done _tr_template="upgrade-${_tr_action:-${_tr_template:?}}.xsl" if [ ! -f "${_tr_schema_o:?}" ] || [ ! -f "${_tr_schema_t:?}" ]; then emit_error "Origin and/or target schema missing, rerun make" return 1 fi while read _tr_origin; do printf '%-60s' "${_tr_origin}... " # pre-validate if ! test_runner_validate "${_tr_schema_o}" "${_tr_origin}"; then _tr_ret=$((_tr_ret + 1)); echo "E:pre-validate"; continue fi # upgrade if ! _tr_target=$(test_runner_upgrade "${_tr_template}" \ "${_tr_origin}" "${_tr_mode}"); then _tr_ret=$((_tr_ret + 1)); if [ -z "$_tr_target" ]; then break fi echo "E:upgrade" if [ -s "$_tr_target" ]; then echo --- cat "$_tr_target" || : echo --- fi continue fi # post-validate if ! test_runner_validate "${_tr_schema_t}" "${_tr_target}"; then _tr_ret=$((_tr_ret + 1)); echo "E:post-validate"; continue fi echo "OK" done log2_or_0_return ${_tr_ret} } # # particular test variations # -C # -X # stdin: granular test specification(s) if any # test2to3() { _t23_pattern= while read _t23_spec; do _t23_spec=${_t23_spec%.xml} _t23_spec=${_t23_spec%\*} _t23_pattern="${_t23_pattern} -name ${_t23_spec}*.xml -o" done if [ -n "$_t23_pattern" ]; then _t23_pattern="( ${_t23_pattern%-o} )" fi find test-2 -name test-2 -o -type d -prune \ -o -name '*.xml' ${_t23_pattern} -print \ | env LC_ALL=C sort \ | { case " $* " in *\ -C\ *) test_cleaner;; *\ -X\ *) test_explanation -o=2.10;; *) test_runner -o=2.10 -t=3.0 "$@" || return $?;; esac; } } test2to3enter() { _t23e_pattern= while read _t23e_spec; do _t23e_spec=${_t23e_spec%.xml} _t23e_spec=${_t23e_spec%\*} _t23e_pattern="${_t23e_pattern} -name ${_t23e_spec}*.xml -o" done if [ -n "$_t23e_pattern" ]; then _t23e_pattern="( ${_t23e_pattern%-o} )" fi find test-2-enter -name test-2-enter -o -type d -prune \ -o -name '*.xml' ${_t23e_pattern} -print \ | env LC_ALL=C sort \ | { case " $* " in *\ -C\ *) test_cleaner;; *\ -X\ *) emit_result "not implemented" "option -X";; *) test_runner -a=2.10-enter -o=2.10 -t=2.10 "$@" || return $?;; esac; } } test2to3leave() { _t23l_pattern= while read _t23l_spec; do _t23l_spec=${_t23l_spec%.xml} _t23l_spec=${_t23l_spec%\*} _t23l_pattern="${_t23l_pattern} -name ${_t23l_spec}*.xml -o" done if [ -n "$_t23l_pattern" ]; then _t23l_pattern="( ${_t23l_pattern%-o} )" fi find test-2-leave -name test-2-leave -o -type d -prune \ -o -name '*.xml' ${_t23l_pattern} -print \ | env LC_ALL=C sort \ | { case " $* " in *\ -C\ *) test_cleaner;; *\ -X\ *) emit_result "not implemented" "option -X";; *) test_runner -a=2.10-leave -o=3.0 -t=3.0 "$@" || return $?;; esac; } } test2to3roundtrip() { _t23rt_pattern= while read _t23tr_spec; do _t23rt_spec=${_t23rt_spec%.xml} _t23rt_spec=${_t23rt_spec%\*} _t23rt_pattern="${_t23rt_pattern} -name ${_t23rt_spec}*.xml -o" done if [ -n "$_t23rt_pattern" ]; then _t23rt_pattern="( ${_t23rt_pattern%-o} )" fi find test-2-roundtrip -name test-2-roundtrip -o -type d -prune \ -o -name '*.xml' ${_t23rt_pattern} -print \ | env LC_ALL=C sort \ | { case " $* " in *\ -C\ *) test_cleaner;; *\ -X\ *) emit_result "not implemented" "option -X";; *) test_runner -a=2.10-roundtrip -o=2.10 -t=3.0 "$@" \ || return $?;; esac; } } # -B # -D # -G ... see usage cts_scheduler() { _tcp_mode=0 _tcp_ret=0 _tcp_validatewith= _tcp_schema_o= _tcp_schema_t= _tcp_template= find ../cts/scheduler/xml -name '*.xml' | env LC_ALL=C sort \ | { case " $* " in *\ -C\ *) test_cleaner -r;; *\ -X\ *) emit_result "not implemented" "option -X";; *) while [ $# -gt 0 ]; do case "$1" in -G) _tcp_mode=$((_tcp_mode | (1 << 0)));; -D) _tcp_mode=$((_tcp_mode | (1 << 1)));; -B) _tcp_mode=$((_tcp_mode | (1 << 2)));; esac shift done while read _tcp_origin; do _tcp_validatewith=$(${XSLT_PROCESSOR} - "${_tcp_origin}" < EOF ) _tcp_schema_t=${_tcp_validatewith} case "${_tcp_validatewith}" in 1) _tcp_schema_o=1.3;; 2) _tcp_schema_o=2.10;; # only for gradual refinement as upgrade-2.10.xsl under # active development, move to 3.x when schema v4 emerges 3) _tcp_schema_o=2.10 _tcp_schema_t=2;; *) emit_error \ "need to skip ${_tcp_origin} (schema: ${_tcp_validatewith})" continue;; esac _tcp_template="upgrade-${_tcp_schema_o}.xsl" _tcp_schema_t="pacemaker-$((_tcp_schema_t + 1)).0.rng" if [ "${_tcp_schema_o%%.*}" = "${_tcp_validatewith}" ]; then _tcp_schema_o="pacemaker-${_tcp_schema_o}.rng" else _tcp_schema_o="${_tcp_schema_t}" fi # pre-validate if [ "$_tcp_schema_o" != "$_tcp_schema_t" ] \ && ! test_runner_validate "$_tcp_schema_o" "$_tcp_origin"; then _tcp_ret=$((_tcp_ret + 1)) echo "E:pre-validate" continue fi # upgrade if [ "$((_tcp_mode & (1 << 0)))" -eq 0 ]; then ln -fs "$(pwd)/$_tcp_origin" "${_tcp_origin%.*}.ref" fi if ! _tcp_target=$(test_runner_upgrade "${_tcp_template}" \ "${_tcp_origin}" "${_tcp_mode}"); then _tcp_ret=$((_tcp_ret + 1)); if [ -z "$_tcp_target" ]; then break fi echo "E:upgrade" if [ -s "$_tcp_target" ]; then echo --- cat "$_tcp_target" || : echo --- fi continue fi if [ "$((_tcp_mode & (1 << 0)))" -eq 0 ]; then rm -f "${_tcp_origin%.*}.ref" fi # post-validate if ! test_runner_validate "${_tcp_schema_t}" "${_tcp_target}"; then _tcp_ret=$((_tcp_ret + 1)); echo "E:post-validate"; continue fi if [ "$((_tcp_mode & (1 << 0)))" -ne 0 ]; then mv "$_tcp_target" "$_tcp_origin" fi done; log2_or_0_return ${_tcp_ret};; esac; } } # # "framework" # # option-likes ... options to be passed down # argument-likes ... drives a test selection test_suite() { _ts_pass= _ts_select= _ts_select_full= _ts_test_specs= _ts_global_ret=0 _ts_ret=0 while [ $# -gt 0 ]; do case "$1" in -) printf '%s\n' 'waiting for tests specified at stdin...'; while read _ts_spec; do _ts_select="${_ts_spec}@$1" done;; -*) _ts_pass="${_ts_pass} $1";; *) _ts_select_full="${_ts_select_full}@$1" _ts_select="${_ts_select}@${1%%/*}";; esac shift done _ts_select="${_ts_select}@" _ts_select_full="${_ts_select_full}@" for _ts_test in ${tests}; do _ts_test_specs= while true; do case "${_ts_select}" in *@${_ts_test}@*) _ts_test_specs="${_ts_select%%@${_ts_test}@*}"\ "@${_ts_select#*@${_ts_test}@}" if [ "$_ts_test_specs" = "@" ]; then _ts_select= # nothing left else _ts_select="${_ts_test_specs}" fi continue ;; @) case "${_ts_test}" in test*) break;; esac # filter ;; esac if [ -n "$_ts_test_specs" ]; then break fi continue 2 # move on to matching with next local test done _ts_test_specs= while true; do case "${_ts_select_full}" in *@${_ts_test}/*) _ts_test_full="${_ts_test}/${_ts_select_full#*@${_ts_test}/}" _ts_test_full="${_ts_test_full%%@*}" _ts_select_full="${_ts_select_full%%@${_ts_test_full}@*}"\ "@${_ts_select_full#*@${_ts_test_full}@}" _ts_test_specs="${_ts_test_specs} ${_ts_test_full#*/}" ;; *) break ;; esac done for _ts_test_spec in ${_ts_test_specs}; do printf '%s\n' "${_ts_test_spec}" done | "${_ts_test}" ${_ts_pass} || _ts_ret=$? if [ "$_ts_ret" = 0 ]; then emit_result "$_ts_ret" "$_ts_test" else emit_result "at least 2^$((_ts_ret - 1))" "$_ts_test" fi log2_or_0_add ${_ts_global_ret} ${_ts_ret} _ts_global_ret=$? done if [ -n "${_ts_select#@}" ]; then emit_error "Non-existing test(s):$(echo "${_ts_select}" \ | tr '@' ' ')" log2_or_0_add ${_ts_global_ret} 1 || _ts_global_ret=$? fi return ${_ts_global_ret} } # NOTE: big letters are dedicated for per-test-set behaviour, # small ones for generic/global behaviour usage() { printf \ '%s\n%s\n %s\n %s\n %s\n %s\n %s\n %s\n %s\n %s\n' \ "usage: $0 [-{B,C,D,G,X}]* \\" \ " [-|{${tests## }}*]" \ "- when no suites (arguments) provided, \"test*\" ones get used" \ "- with '-' suite specification the actual ones grabbed on stdin" \ "- use '-B' to run validate-only check suppressing blanks first" \ "- use '-C' to only cleanup ephemeral byproducts" \ "- use '-D' to review originals vs. \"referential\" outcomes" \ "- use '-G' to generate \"referential\" outcomes" \ "- use '-X' to show explanatory details about the upgrade" \ "- test specification can be granular, e.g. 'test2to3/022'" } main() { _main_pass= _main_bailout=0 _main_ret=0 while [ $# -gt 0 ]; do case "$1" in -h) usage; exit;; -C|-G|-X) _main_bailout=1;; esac _main_pass="${_main_pass} $1" shift done test_suite ${_main_pass} || _main_ret=$? if [ "$_main_bailout" -eq 0 ]; then test_suite -C $_main_pass >/dev/null || true fi if [ "$_main_ret" = 0 ]; then emit_result "$_main_ret" "Overall suite" else emit_result "at least 2^$((_main_ret - 1))" "Overall suite" fi return ${_main_ret} } main "$@"