haproxy/scripts/git-show-backports
Willy Tarreau f97db23b6d SCRIPTS: git-show-backports: automatic ref and base detection with -m
When running with -m (check for missing backports) we often have to fill
lots of information that can be determined automatically the vast majority
of the time:
  - restart point (last cherry-picked ID from one of the last commits)
  - current branch (HEAD)
  - reference branch (the one that contains most of the last commits)

These elements are not that hard to determine, so let's make sure we
can fall back to them when running in missing mode.

The reference branch is guessed by looking at the upstream branch that
most frequently contains some of the last 10 commits. It can be inaccurate
if multiple branches exist with these commits, or when upstream changes
due to a non-LTS branch disappearing in the middle of the series, in which
case passing "-r" will help. But most of the time it works OK. It also gives
precedence to local branches over remote ones for such choices. A test in
2.4 at commit 793a4b520 correctly shows 2.6/master as the upstream despite
2.5 having been used for the early ones of the tag.

For the restart point, we assume that the most recent commit that was
backported serves as a reference (and not the most recently backported
commit). This means that the usual case where an old commit was found
to be missing will not fool the analysis. Commits are inspected from
2 commits before the last tag, and reordered from the parent's tree
to see which one is the last one.

With this, it's sufficient to issue "git-show-backports -q -m" to get
the list of backports from the upstream branch, restarting from the
last backported one.
2023-08-14 13:12:56 +02:00

337 lines
10 KiB
Bash
Executable File

#!/usr/bin/env bash
#
# Compares multiple branches against a reference and shows which ones contain
# each commit, and the level of backports since the origin or its own ancestors.
#
# Copyright (c) 2016 Willy Tarreau <w@1wt.eu>
#
# The purpose is to make it easy to visualize what backports might be missing
# in a maintenance branch, and to easily spot the ones that are needed and the
# ones that are not. It solely relies on the "cherry-picked from" tags in the
# commit messages to find what commit is available where, and can even find a
# reference commit's ancestor in another branch's commit ancestors as well to
# detect that the patch is present. When done with the proper references and
# a correct ordering of the branches, it can be used to quickly apply a set of
# fixes to a branch since it dumps suggested commands at the end. When doing
# so it is a good idea to use "HEAD" as the last branch to avoid doing mistakes.
#
# Examples :
# - find what's in master and not in current branch :
# show-backports -q -m -r master HEAD
# - find what's in 1.6/master and in hapee-maint-1.5r2 but not in current branch :
# show-backports -q -m -r 1.6/master hapee-maint-1.5r2 HEAD | grep ' [a-f0-9]\{8\}[-+][0-9] '
# - check that no recent fix from master is missing in any maintenance branch :
# show-backports -r master hapee-maint-1.5r2 aloha-7.5 hapee-maint-1.5r1 aloha-7.0
# - see what was recently merged into 1.6 and has no equivalent in local master :
# show-backports -q -m -r 1.6/master -b "1.6/master@{1 week ago}" master
# - check what extra backports are present in hapee-r2 compared to hapee-r1 :
# show-backports -q -m -r hapee-r2 hapee-r1
USAGE="Usage: ${0##*/} [-q] [-H] [-m] [-u] [-r reference] [-l logexpr] [-s subject] [-b base] {branch|range} [...] [-- file*]"
BASES=( )
BRANCHES=( )
REF=
BASE=
QUIET=
LOGEXPR=
SUBJECT=
MISSING=
UPSTREAM=
BODYHASH=
die() {
[ "$#" -eq 0 ] || echo "$*" >&2
exit 1
}
err() {
echo "$*" >&2
}
quit() {
[ "$#" -eq 0 ] || echo "$*"
exit 0
}
short() {
# git rev-parse --short $1
echo "${1::8}"
}
dump_commit_matrix() {
title=":$REF:"
for branch in "${BRANCHES[@]}"; do
#echo -n " $branch"
title="$title :${branch}:"
done
title="$title |"
count=0
# now look up commits
while read ref subject; do
if [ -n "$MISSING" -a "${subject:0:9}" = "[RELEASE]" ]; then
continue
fi
upstream="none"
missing=0
refbhash=""
line=""
for branch in "${BRANCHES[@]}"; do
set -- $(grep -m 1 $ref "$WORK/${branch//\//_}")
newhash=$1 ; shift
bhash=""
# count the number of cherry-picks after this one. Since we shift,
# the result is in "$#"
while [ -n "$1" -a "$1" != "$ref" ]; do
shift
done
if [ -n "$newhash" ]; then
line="${line} $(short $newhash)-$#"
else
# before giving up we can check if our current commit was
# itself cherry-picked and check this again. In order not
# to have to do it all the time, we can cache the result
# for the current line. If a match is found we report it
# with the '+' delimiter instead of '-'.
if [ "$upstream" = "none" ]; then
upstream=( $(git log -1 --pretty --format=%B "$ref" | \
sed -n 's/^commit \([^)]*\) upstream\.$/\1/p;s/^(cherry picked from commit \([^)]*\))/\1/p') )
fi
newhash=""
for h in ${upstream[@]}; do
set -- $(grep -m 1 $h "$WORK/${branch//\//_}")
newhash=$1 ; shift
while [ -n "$1" -a "$1" != "$h" ]; do
shift
done
if [ -n "$newhash" ]; then
line="${line} $(short $newhash)+$#"
break
fi
done
if [ -z "$newhash" -a -n "$BODYHASH" ]; then
if [ -z "$refbhash" ]; then
refbhash=$(git log -1 --pretty="%an|%ae|%at|%B" "$ref" | sed -n '/^\(Signed-off-by\|(cherry picked\)/q;p' | md5sum)
fi
set -- $(grep -m 1 "H$refbhash\$" "$WORK/${branch//\//_}")
newhash=$1 ; shift
if [ -n "$newhash" ]; then
line="${line} $(short $newhash)+?"
break
fi
fi
if [ -z "$newhash" ]; then
line="${line} -"
missing=1
fi
fi
done
line="${line} |"
if [ -z "$MISSING" -o $missing -gt 0 ]; then
[ $((count++)) -gt 0 ] || echo "$title"
[ "$QUIET" != "" -o $count -lt 20 ] || count=0
if [ -z "$UPSTREAM" -o "$upstream" = "none" -o -z "$upstream" ]; then
echo "$(short $ref) $line"
else
echo "$(short $upstream) $line"
fi
fi
done < "$WORK/${REF//\//_}"
}
while [ -n "$1" -a -z "${1##-*}" ]; do
case "$1" in
-b) BASE="$2" ; shift 2 ;;
-r) REF="$2" ; shift 2 ;;
-l) LOGEXPR="$2" ; shift 2 ;;
-s) SUBJECT="$2" ; shift 2 ;;
-q) QUIET=1 ; shift ;;
-m) MISSING=1 ; shift ;;
-u) UPSTREAM=1 ; shift ;;
-H) BODYHASH=1 ; shift ;;
-h|--help) quit "$USAGE" ;;
*) die "$USAGE" ;;
esac
done
# if no ref, either we're checking missing backports and we'll guess
# the upstream reference branch based on which one contains most of
# the latest commits, or we'll use master.
if [ -z "$REF" ]; then
if [ -n "$MISSING" ]; then
# check the last 10 commits in the base branch, and see where
# the seem to be coming from.
TAG="$(git describe --tags ${BASE:-HEAD} --abbrev=0)"
LAST_COMMITS=( $(git rev-list --abbrev-commit --reverse "$TAG^^.." | tail -n10) )
REF=$(for i in "${LAST_COMMITS[@]}"; do
upstream=$(git log -1 --pretty --format=%B $i |
sed -n 's/^commit \([^)]*\) upstream\.$/\1/p;s/^(cherry picked from commit \([^)]*\))/\1/p' |
tail -n1)
if [ -n "$upstream" ]; then
# use local first then remote branch
( git branch --sort=refname --contains $upstream | head -n1 ;
git branch -r --sort=refname --contains $upstream | head -n1) 2>&1 |
grep 'master\|maint' | head -n1
fi
done | sort | uniq -c | sort -nr | awk '{ print $NF; exit;}')
# here we have a name, e.g. "2.6/master" in REF
REF="${REF:-master}"
err "Warning! No ref specified, using $REF."
else
REF=master
fi
fi
# branches may also appear as id1..id2 to limit the history instead of looking
# back to the common base. The field is left empty if not set.
BRANCHES=( )
BASES=( )
while [ $# -gt 0 ]; do
if [ "$1" = "--" ]; then
shift
break
fi
branch="${1##*..}"
if [ "$branch" == "$1" ]; then
base=""
else
base="${1%%..*}"
fi
BASES[${#BRANCHES[@]}]="$base"
BRANCHES[${#BRANCHES[@]}]="$branch"
shift
done
# args left for git-log
ARGS=( "$@" )
if [ ${#BRANCHES[@]} = 0 ]; then
if [ -n "$MISSING" ]; then
BRANCHES=( HEAD )
else
die "$USAGE"
fi
fi
for branch in "$REF" "${BRANCHES[@]}"; do
if ! git rev-parse --verify -q "$branch" >/dev/null; then
die "Failed to check git branch $branch."
fi
done
if [ -z "$BASE" -a -n "$MISSING" ]; then
err "Warning! No base specified, checking latest backports from current branch since last tag."
TAG="$(git describe --tags HEAD --abbrev=0)"
COMMITS=( $(git rev-list --abbrev-commit --reverse "$TAG^^..") )
tip=""
for commit in "${COMMITS[@]}"; do
parent=$(git log -1 --pretty --format=%B $commit |
sed -n 's/^commit \([^)]*\) upstream\.$/\1/p;s/^(cherry picked from commit \([^)]*\))/\1/p' |
tail -n1)
if [ -z "$tip" ]; then
tip=$parent
elif [ -n "$parent" ]; then
base=$(git merge-base "$tip" "$parent")
if [ "$base" = "$tip" ]; then
# tip is older than parent, switch tip to it if it
# belongs to the upstream branch
if [ "$(git merge-base $parent $REF)" = "$parent" ]; then
tip=$parent
fi
fi
fi
done
BASE="$tip"
if [ -n "$BASE" ]; then
echo "Restarting from $(git log -1 --no-decorate --oneline $BASE)"
else
echo "Could not figure the base."
fi
fi
if [ -z "$BASE" ]; then
err "Warning! No base specified, looking for common ancestor."
BASE=$(git merge-base --all "$REF" "${BRANCHES[@]}")
if [ -z "$BASE" ]; then
die "Couldn't find a common ancestor between these branches"
fi
fi
# we want to go to the git root dir
DIR="$PWD"
cd $(git rev-parse --show-toplevel)
mkdir -p .git/.show-backports #|| die "Can't create .git/.show-backports"
WORK=.git/.show-backports
rm -f "$WORK/${REF//\//_}"
git log --reverse ${LOGEXPR:+--grep $LOGEXPR} --pretty="%H %s" "$BASE".."$REF" -- "${ARGS[@]}" | grep "${SUBJECT}" > "$WORK/${REF//\//_}"
# for each branch, enumerate all commits and their ancestry
branch_num=0;
while [ $branch_num -lt "${#BRANCHES[@]}" ]; do
branch="${BRANCHES[$branch_num]}"
base="${BASES[$branch_num]}"
base="${base:-$BASE}"
rm -f "$WORK/${branch//\//_}"
git log --reverse --pretty="%H %s" "$base".."$branch" -- "${ARGS[@]}" | grep "${SUBJECT}" | while read h subject; do
echo -n "$h" $(git log -1 --pretty --format=%B "$h" | \
sed -n 's/^commit \([^)]*\) upstream\.$/\1/p;s/^(cherry picked from commit \([^)]*\))/\1/p')
if [ -n "$BODYHASH" ]; then
echo " H$(git log -1 --pretty="%an|%ae|%at|%B" "$h" | sed -n '/^\(Signed-off-by\|(cherry picked\)/q;p' | md5sum)"
else
echo
fi
done > "$WORK/${branch//\//_}"
(( branch_num++ ))
done
count=0
dump_commit_matrix | column -t | \
(
left_commits=( )
right_commits=( )
while read line; do
# append the subject at the end of the line
set -- $line
echo -n "$line "
if [ "${line::1}" = ":" ]; then
echo "---- Subject ----"
else
# doing it this way prevents git from abusing the terminal
echo "$(git log -1 --pretty="%s" "$1")"
left_commits[${#left_commits[@]}]="$1"
comm=""
while [ -n "$1" -a "$1" != "-" -a "$1" != "|" ]; do
comm="${1%-*}"
shift
done
right_commits[${#right_commits[@]}]="$comm"
fi
done
if [ -n "$MISSING" -a ${#left_commits[@]} -eq 0 ]; then
echo "No missing commit to apply."
elif [ -n "$MISSING" ]; then
echo
echo
echo "In order to show and/or apply all leftmost commits to current branch :"
echo " git show --pretty=format:'%C(yellow)commit %H%C(normal)%nAuthor: %an <%ae>%nDate: %aD%n%n%C(green)%C(bold)git cherry-pick -sx %h%n%n%w(72,4,4)%B%N' ${left_commits[@]}"
echo
echo " git cherry-pick -sx ${left_commits[@]}"
echo
if [ "${left_commits[*]}" != "${right_commits[*]}" ]; then
echo "In order to show and/or apply all rightmost commits to current branch :"
echo " git show --pretty=format:'%C(yellow)commit %H%C(normal)%nAuthor: %an <%ae>%nDate: %aD%n%n%C(green)%C(bold)git cherry-pick -sx %h%n%n%w(72,4,4)%B%N' ${right_commits[@]}"
echo
echo " git cherry-pick -sx ${right_commits[@]}"
echo
fi
fi
)