buildroot/utils/test-pkg

270 lines
8.1 KiB
Plaintext
Raw Normal View History

#!/usr/bin/env bash
set -e
TOOLCHAINS_CSV='support/config-fragments/autobuild/toolchain-configs.csv'
TEMP_CONF=""
do_clean() {
if [ ! -z "${TEMP_CONF}" ]; then
rm -f "${TEMP_CONF}"
fi
}
main() {
local o O opts
local cfg dir pkg random toolchains_csv toolchain all number mode
local ret nb nb_skip nb_fail nb_legal nb_tc build_dir keep
local -a toolchains
local pkg_br_name
o='hakc:d:n:p:r:t:'
O='help,all,keep,config-snippet:,build-dir:,number:,package:,random:,toolchains-csv:'
opts="$(getopt -n "${my_name}" -o "${o}" -l "${O}" -- "${@}")"
eval set -- "${opts}"
random=0
test-pkg: test a subset of toolchains by default, add -a and -n options During the latest Buildroot Developers meeting, we discussed that test-pkg would perhaps be more widely used if it tested a smaller subset of toolchains. Indeed, it currently tests 47 toolchains, which takes very long to build. Several of the toolchain configurations are quite similar, and it is perhaps not necessary for contributors to test them all before submitting a package. Therefore, this commit changes the test-pkg script to only test a subset of the toolchain configurations by default. The N first configurations of the CSV files are tested, where N is hard-coded in the script. The CSV file has therefore been re-organized to have the first N toolchains be the most important ones. A -a/--all option is added to test with all toolchains, while a -n/--number option is added to test with the first N toolchains, N being passed on the command line. Note that the list of toolchains (built in the "toolchains" shell variable) is no longer sorted. Indeed, when the first N toolchains are tested, we want them to be tested in the same order as they are listed in the CSV file, as we are careful to order them in an interesting order. We only sort when all toolchains are tested. Signed-off-by: Thomas Petazzoni <thomas.petazzoni@free-electrons.com> Tested-by: Matt Weber <matthew.weber@rockwellcollins.com> Reviewed-by: "Yann E. MORIN" <yann.morin.1998@free.fr> Signed-off-by: Thomas Petazzoni <thomas.petazzoni@bootlin.com>
2018-03-24 05:48:14 +08:00
all=0
keep=0
test-pkg: test a subset of toolchains by default, add -a and -n options During the latest Buildroot Developers meeting, we discussed that test-pkg would perhaps be more widely used if it tested a smaller subset of toolchains. Indeed, it currently tests 47 toolchains, which takes very long to build. Several of the toolchain configurations are quite similar, and it is perhaps not necessary for contributors to test them all before submitting a package. Therefore, this commit changes the test-pkg script to only test a subset of the toolchain configurations by default. The N first configurations of the CSV files are tested, where N is hard-coded in the script. The CSV file has therefore been re-organized to have the first N toolchains be the most important ones. A -a/--all option is added to test with all toolchains, while a -n/--number option is added to test with the first N toolchains, N being passed on the command line. Note that the list of toolchains (built in the "toolchains" shell variable) is no longer sorted. Indeed, when the first N toolchains are tested, we want them to be tested in the same order as they are listed in the CSV file, as we are careful to order them in an interesting order. We only sort when all toolchains are tested. Signed-off-by: Thomas Petazzoni <thomas.petazzoni@free-electrons.com> Tested-by: Matt Weber <matthew.weber@rockwellcollins.com> Reviewed-by: "Yann E. MORIN" <yann.morin.1998@free.fr> Signed-off-by: Thomas Petazzoni <thomas.petazzoni@bootlin.com>
2018-03-24 05:48:14 +08:00
number=0
mode=0
toolchains_csv="${TOOLCHAINS_CSV}"
while [ ${#} -gt 0 ]; do
case "${1}" in
(-h|--help)
help; exit 0
;;
test-pkg: test a subset of toolchains by default, add -a and -n options During the latest Buildroot Developers meeting, we discussed that test-pkg would perhaps be more widely used if it tested a smaller subset of toolchains. Indeed, it currently tests 47 toolchains, which takes very long to build. Several of the toolchain configurations are quite similar, and it is perhaps not necessary for contributors to test them all before submitting a package. Therefore, this commit changes the test-pkg script to only test a subset of the toolchain configurations by default. The N first configurations of the CSV files are tested, where N is hard-coded in the script. The CSV file has therefore been re-organized to have the first N toolchains be the most important ones. A -a/--all option is added to test with all toolchains, while a -n/--number option is added to test with the first N toolchains, N being passed on the command line. Note that the list of toolchains (built in the "toolchains" shell variable) is no longer sorted. Indeed, when the first N toolchains are tested, we want them to be tested in the same order as they are listed in the CSV file, as we are careful to order them in an interesting order. We only sort when all toolchains are tested. Signed-off-by: Thomas Petazzoni <thomas.petazzoni@free-electrons.com> Tested-by: Matt Weber <matthew.weber@rockwellcollins.com> Reviewed-by: "Yann E. MORIN" <yann.morin.1998@free.fr> Signed-off-by: Thomas Petazzoni <thomas.petazzoni@bootlin.com>
2018-03-24 05:48:14 +08:00
(-a|--all)
all=1; shift 1
;;
(-k|--keep)
keep=1; shift 1
;;
(-c|--config-snippet)
cfg="${2}"; shift 2
;;
(-d|--build-dir)
dir="${2}"; shift 2
;;
test-pkg: test a subset of toolchains by default, add -a and -n options During the latest Buildroot Developers meeting, we discussed that test-pkg would perhaps be more widely used if it tested a smaller subset of toolchains. Indeed, it currently tests 47 toolchains, which takes very long to build. Several of the toolchain configurations are quite similar, and it is perhaps not necessary for contributors to test them all before submitting a package. Therefore, this commit changes the test-pkg script to only test a subset of the toolchain configurations by default. The N first configurations of the CSV files are tested, where N is hard-coded in the script. The CSV file has therefore been re-organized to have the first N toolchains be the most important ones. A -a/--all option is added to test with all toolchains, while a -n/--number option is added to test with the first N toolchains, N being passed on the command line. Note that the list of toolchains (built in the "toolchains" shell variable) is no longer sorted. Indeed, when the first N toolchains are tested, we want them to be tested in the same order as they are listed in the CSV file, as we are careful to order them in an interesting order. We only sort when all toolchains are tested. Signed-off-by: Thomas Petazzoni <thomas.petazzoni@free-electrons.com> Tested-by: Matt Weber <matthew.weber@rockwellcollins.com> Reviewed-by: "Yann E. MORIN" <yann.morin.1998@free.fr> Signed-off-by: Thomas Petazzoni <thomas.petazzoni@bootlin.com>
2018-03-24 05:48:14 +08:00
(-n|--number)
number="${2}"; shift 2
;;
(-p|--package)
pkg="${2}"; shift 2
;;
(-r|--random)
random="${2}"; shift 2
;;
(-t|--toolchains-csv)
toolchains_csv="${2}"; shift 2
;;
(--)
shift; break
;;
esac
done
trap do_clean INT TERM HUP EXIT
if [ -z "${cfg}" ]; then
pkg_br_name="${pkg//-/_}"
pkg_br_name="BR2_PACKAGE_${pkg_br_name^^}"
TEMP_CONF=$(mktemp /tmp/test-${pkg}-config.XXXXXX)
echo "${pkg_br_name}=y" > ${TEMP_CONF}
cfg="${TEMP_CONF}"
fi
if [ ! -e "${cfg}" ]; then
printf "error: %s: no such file\n" "${cfg}" >&2; exit 1
fi
if [ -z "${dir}" ]; then
dir="${HOME}/br-test-pkg"
fi
test-pkg: test a subset of toolchains by default, add -a and -n options During the latest Buildroot Developers meeting, we discussed that test-pkg would perhaps be more widely used if it tested a smaller subset of toolchains. Indeed, it currently tests 47 toolchains, which takes very long to build. Several of the toolchain configurations are quite similar, and it is perhaps not necessary for contributors to test them all before submitting a package. Therefore, this commit changes the test-pkg script to only test a subset of the toolchain configurations by default. The N first configurations of the CSV files are tested, where N is hard-coded in the script. The CSV file has therefore been re-organized to have the first N toolchains be the most important ones. A -a/--all option is added to test with all toolchains, while a -n/--number option is added to test with the first N toolchains, N being passed on the command line. Note that the list of toolchains (built in the "toolchains" shell variable) is no longer sorted. Indeed, when the first N toolchains are tested, we want them to be tested in the same order as they are listed in the CSV file, as we are careful to order them in an interesting order. We only sort when all toolchains are tested. Signed-off-by: Thomas Petazzoni <thomas.petazzoni@free-electrons.com> Tested-by: Matt Weber <matthew.weber@rockwellcollins.com> Reviewed-by: "Yann E. MORIN" <yann.morin.1998@free.fr> Signed-off-by: Thomas Petazzoni <thomas.petazzoni@bootlin.com>
2018-03-24 05:48:14 +08:00
if [ ${random} -gt 0 ]; then
mode=$((mode+1))
fi
if [ ${number} -gt 0 ]; then
mode=$((mode+1))
fi
if [ ${all} -eq 1 ]; then
mode=$((mode+1))
fi
# Default mode is to test the N first toolchains, which have been
# chosen to be a good selection of toolchains.
if [ ${mode} -eq 0 ] ; then
number=6
test-pkg: test a subset of toolchains by default, add -a and -n options During the latest Buildroot Developers meeting, we discussed that test-pkg would perhaps be more widely used if it tested a smaller subset of toolchains. Indeed, it currently tests 47 toolchains, which takes very long to build. Several of the toolchain configurations are quite similar, and it is perhaps not necessary for contributors to test them all before submitting a package. Therefore, this commit changes the test-pkg script to only test a subset of the toolchain configurations by default. The N first configurations of the CSV files are tested, where N is hard-coded in the script. The CSV file has therefore been re-organized to have the first N toolchains be the most important ones. A -a/--all option is added to test with all toolchains, while a -n/--number option is added to test with the first N toolchains, N being passed on the command line. Note that the list of toolchains (built in the "toolchains" shell variable) is no longer sorted. Indeed, when the first N toolchains are tested, we want them to be tested in the same order as they are listed in the CSV file, as we are careful to order them in an interesting order. We only sort when all toolchains are tested. Signed-off-by: Thomas Petazzoni <thomas.petazzoni@free-electrons.com> Tested-by: Matt Weber <matthew.weber@rockwellcollins.com> Reviewed-by: "Yann E. MORIN" <yann.morin.1998@free.fr> Signed-off-by: Thomas Petazzoni <thomas.petazzoni@bootlin.com>
2018-03-24 05:48:14 +08:00
elif [ ${mode} -gt 1 ] ; then
printf "error: --all, --number and --random are mutually exclusive\n" >&2; exit 1
fi
# Extract the URLs of the toolchains; drop internal toolchains
# E.g.: http://server/path/to/name.config,arch,libc
# --> http://server/path/to/name.config
toolchains=($(sed -r -e 's/,.*//; /internal/d; /^#/d; /^$/d;' "${toolchains_csv}" \
|if [ ${random} -gt 0 ]; then \
sort -R |head -n ${random}
test-pkg: test a subset of toolchains by default, add -a and -n options During the latest Buildroot Developers meeting, we discussed that test-pkg would perhaps be more widely used if it tested a smaller subset of toolchains. Indeed, it currently tests 47 toolchains, which takes very long to build. Several of the toolchain configurations are quite similar, and it is perhaps not necessary for contributors to test them all before submitting a package. Therefore, this commit changes the test-pkg script to only test a subset of the toolchain configurations by default. The N first configurations of the CSV files are tested, where N is hard-coded in the script. The CSV file has therefore been re-organized to have the first N toolchains be the most important ones. A -a/--all option is added to test with all toolchains, while a -n/--number option is added to test with the first N toolchains, N being passed on the command line. Note that the list of toolchains (built in the "toolchains" shell variable) is no longer sorted. Indeed, when the first N toolchains are tested, we want them to be tested in the same order as they are listed in the CSV file, as we are careful to order them in an interesting order. We only sort when all toolchains are tested. Signed-off-by: Thomas Petazzoni <thomas.petazzoni@free-electrons.com> Tested-by: Matt Weber <matthew.weber@rockwellcollins.com> Reviewed-by: "Yann E. MORIN" <yann.morin.1998@free.fr> Signed-off-by: Thomas Petazzoni <thomas.petazzoni@bootlin.com>
2018-03-24 05:48:14 +08:00
elif [ ${number} -gt 0 ]; then \
head -n ${number}
else
sort
fi
)
)
nb_tc="${#toolchains[@]}"
if [ ${nb_tc} -eq 0 ]; then
printf "error: no toolchain found (networking issue?)\n" >&2; exit 1
fi
nb=0
nb_skip=0
nb_fail=0
nb_legal=0
for toolchainconfig in "${toolchains[@]}"; do
: $((nb++))
toolchain="$(basename "${toolchainconfig}" .config)"
build_dir="${dir}/${toolchain}"
printf "%40s [%*d/%d]: " "${toolchain}" ${#nb_tc} ${nb} ${nb_tc}
build_one "${build_dir}" "${toolchainconfig}" "${cfg}" "${pkg}" && ret=0 || ret=${?}
case ${ret} in
(0) printf "OK\n";;
(1) : $((nb_skip++)); printf "SKIPPED\n";;
(2) : $((nb_fail++)); printf "FAILED\n";;
(3) : $((nb_legal++)); printf "FAILED\n";;
esac
done
printf "%d builds, %d skipped, %d build failed, %d legal-info failed\n" \
${nb} ${nb_skip} ${nb_fail} ${nb_legal}
return $((nb_fail + nb_legal))
}
build_one() {
local dir="${1}"
local toolchainconfig="${2}"
local cfg="${3}"
local pkg="${4}"
mkdir -p "${dir}"
CONFIG_= support/kconfig/merge_config.sh -O "${dir}" \
"${toolchainconfig}" "support/config-fragments/minimal.config" "${cfg}" \
>> "${dir}/logfile" 2>&1
# We want all the options from the snippet to be present as-is (set
# or not set) in the actual .config; if one of them is not, it means
# some dependency from the toolchain or arch is not available, in
# which case this config is untestable and we skip it.
# We don't care about the locale to sort in, as long as both sort are
# done in the same locale.
comm -23 <(sort "${cfg}") <(sort "${dir}/.config") >"${dir}/missing.config"
if [ -s "${dir}/missing.config" ]; then
return 1
fi
# Remove file, it's empty anyway.
rm -f "${dir}/missing.config"
if [ -n "${pkg}" ]; then
if ! make O="${dir}" "${pkg}-dirclean" >> "${dir}/logfile" 2>&1; then
return 2
fi
fi
# shellcheck disable=SC2086
if ! BR_FORCE_CHECK_DEPENDENCIES=YES make O="${dir}" ${pkg} >> "${dir}/logfile" 2>&1; then
return 2
fi
# legal-info done systematically, because some packages have different
# sources depending on the configuration (e.g. lua-5.2 vs. lua-5.3)
if ! make O="${dir}" legal-info >> "${dir}/logfile" 2>&1; then
return 3
fi
# If we get here, the build was successful. Clean up the build/host
# directories to save disk space, unless 'keep' was set.
if [ ${keep} -ne 1 ]; then
make O="${dir}" clean >> "${dir}/logfile" 2>&1
fi
}
help() {
cat <<_EOF_
test-pkg: test-build a package against various toolchains and architectures
The supplied config snippet is appended to each toolchain config, the
resulting configuration is checked to ensure it still contains all options
specified in the snippet; if any is missing, the build is skipped, on the
assumption that the package under test requires a toolchain or architecture
feature that is missing.
In case failures are noticed, you can fix the package and just re-run the
same command again; it will re-run the test where it failed. If you did
specify a package (with -p), the package build dir will be removed first.
The list of toolchains is retrieved from ${TOOLCHAINS_CSV}.
Only the external toolchains are tried, because building a Buildroot toolchain
would take too long. An alternative toolchains CSV file can be specified with
the -t option. This file should have lines consisting of the path to the
toolchain config fragment and the required host architecture, separated by a
comma. The config fragments should contain only the toolchain and architecture
settings.
test-pkg: test a subset of toolchains by default, add -a and -n options During the latest Buildroot Developers meeting, we discussed that test-pkg would perhaps be more widely used if it tested a smaller subset of toolchains. Indeed, it currently tests 47 toolchains, which takes very long to build. Several of the toolchain configurations are quite similar, and it is perhaps not necessary for contributors to test them all before submitting a package. Therefore, this commit changes the test-pkg script to only test a subset of the toolchain configurations by default. The N first configurations of the CSV files are tested, where N is hard-coded in the script. The CSV file has therefore been re-organized to have the first N toolchains be the most important ones. A -a/--all option is added to test with all toolchains, while a -n/--number option is added to test with the first N toolchains, N being passed on the command line. Note that the list of toolchains (built in the "toolchains" shell variable) is no longer sorted. Indeed, when the first N toolchains are tested, we want them to be tested in the same order as they are listed in the CSV file, as we are careful to order them in an interesting order. We only sort when all toolchains are tested. Signed-off-by: Thomas Petazzoni <thomas.petazzoni@free-electrons.com> Tested-by: Matt Weber <matthew.weber@rockwellcollins.com> Reviewed-by: "Yann E. MORIN" <yann.morin.1998@free.fr> Signed-off-by: Thomas Petazzoni <thomas.petazzoni@bootlin.com>
2018-03-24 05:48:14 +08:00
By default, a useful subset of toolchains is tested. If needed, all
toolchains can be tested (-a), an arbitrary number of toolchains (-n
in order, -r for random).
Options:
-h, --help
Print this help.
-c CFG, --config-snippet CFG
Use the CFG file as the source for the config snippet. This file
should contain all the config options required to build a package.
-d DIR, --build-dir DIR
Do the builds in directory DIR, one sub-dir per toolchain.
-p PKG, --package PKG
Test-build the package PKG, by running 'make PKG'; if not specified,
just runs 'make'.
test-pkg: test a subset of toolchains by default, add -a and -n options During the latest Buildroot Developers meeting, we discussed that test-pkg would perhaps be more widely used if it tested a smaller subset of toolchains. Indeed, it currently tests 47 toolchains, which takes very long to build. Several of the toolchain configurations are quite similar, and it is perhaps not necessary for contributors to test them all before submitting a package. Therefore, this commit changes the test-pkg script to only test a subset of the toolchain configurations by default. The N first configurations of the CSV files are tested, where N is hard-coded in the script. The CSV file has therefore been re-organized to have the first N toolchains be the most important ones. A -a/--all option is added to test with all toolchains, while a -n/--number option is added to test with the first N toolchains, N being passed on the command line. Note that the list of toolchains (built in the "toolchains" shell variable) is no longer sorted. Indeed, when the first N toolchains are tested, we want them to be tested in the same order as they are listed in the CSV file, as we are careful to order them in an interesting order. We only sort when all toolchains are tested. Signed-off-by: Thomas Petazzoni <thomas.petazzoni@free-electrons.com> Tested-by: Matt Weber <matthew.weber@rockwellcollins.com> Reviewed-by: "Yann E. MORIN" <yann.morin.1998@free.fr> Signed-off-by: Thomas Petazzoni <thomas.petazzoni@bootlin.com>
2018-03-24 05:48:14 +08:00
-a, --all
Test all toolchains, instead of the default subset defined by
Buildroot developers.
-n N, --number N
Test N toolchains, in the order defined in the toolchain CSV
file.
-r N, --random N
test-pkg: test a subset of toolchains by default, add -a and -n options During the latest Buildroot Developers meeting, we discussed that test-pkg would perhaps be more widely used if it tested a smaller subset of toolchains. Indeed, it currently tests 47 toolchains, which takes very long to build. Several of the toolchain configurations are quite similar, and it is perhaps not necessary for contributors to test them all before submitting a package. Therefore, this commit changes the test-pkg script to only test a subset of the toolchain configurations by default. The N first configurations of the CSV files are tested, where N is hard-coded in the script. The CSV file has therefore been re-organized to have the first N toolchains be the most important ones. A -a/--all option is added to test with all toolchains, while a -n/--number option is added to test with the first N toolchains, N being passed on the command line. Note that the list of toolchains (built in the "toolchains" shell variable) is no longer sorted. Indeed, when the first N toolchains are tested, we want them to be tested in the same order as they are listed in the CSV file, as we are careful to order them in an interesting order. We only sort when all toolchains are tested. Signed-off-by: Thomas Petazzoni <thomas.petazzoni@free-electrons.com> Tested-by: Matt Weber <matthew.weber@rockwellcollins.com> Reviewed-by: "Yann E. MORIN" <yann.morin.1998@free.fr> Signed-off-by: Thomas Petazzoni <thomas.petazzoni@bootlin.com>
2018-03-24 05:48:14 +08:00
Limit the tests to the N randomly selected toolchains.
-t CSVFILE, --toolchains-csv CSVFILE
CSV file containing the paths to config fragments of toolchains to
try. If not specified, the toolchains in ${TOOLCHAINS_CSV} will be
used.
-k, --keep
Keep the build directories even if the build succeeds.
Note: the logfile and configuration is always retained, even without
this option.
Example:
Testing libcec would require a config snippet that contains:
BR2_PACKAGE_LIBCEC=y
Testing libcurl with openSSL support would require a snippet such as:
BR2_PACKAGE_OPENSSL=y
BR2_PACKAGE_LIBCURL=y
_EOF_
}
my_name="${0##*/}"
main "${@}"