new script to merge in upstream changes

Author: jason-famedly

merge_upstream.sh

@@ -0,0 +1,197 @@
+#!/bin/bash
+
+set -e
+
+# For a given argument that is a release version, it will be the tag. For now, want the
+# format of `vX.Y.Z_A`. The underscore and `A` is important to include, as that is our
+# patch version that is separate from the upstream patch version
+release_name=$1
+echo "release_name ${release_name}"
+
+# The minor number of the version, so of the release_name above which looks like `vX.Y`,
+# we are peeling off the `Y` for further use
+expected_minor=${release_name##*.}
+echo "expected_minor ${expected_minor}"
+
+# What our branch name will be. Upstream uses the format of `release-vX.Y` so to avoid
+# confusion we will pre-pend 'famedly' to that and use a `/` to namespace it. Like this:
+# `famedly-release/vX.Y`. We do want patch levels and hotfixes to live on the same
+# branch as the release major.minor
+release_branch_name="famedly-release/${release_name}"
+echo "release_branch_name ${release_branch_name}"
+
+upstream_release_branch_name="release-${release_name}"
+echo "upstream_release_branch_name ${upstream_release_branch_name}"
+
+logical_cores=$([ $(uname) = 'Darwin' ] &&
+                       sysctl -n hw.logicalcpu_max ||
+                       nproc)
+
+if [ -z "${release_name}" ]; then
+    echo "Usage: $0 <release_name>"
+    exit 1
+fi
+
+if [ "${release_name}" = "-h" ]; then
+    echo "Usage: $0 <release_name>"
+    exit 0
+fi
+
+echo -e "\e[32m>>>> fetching origin branches\e[0m"
+# If the release_name that was passed already exists as a tag, this will fatal error.
+# Make sure it does not legitimately exist, or that you really mean to replace it before
+# running `git tag -d <release_name>` to remove that one single tag
+git fetch --tags origin
+
+echo -e "\e[32m>>>> fetching upstream branches\e[0m"
+git fetch --tags upstream
+
+echo -e "\e[32m>>>> checkout master branch\e[0m"
+git checkout master-test
+echo -e "\e[32m>>>> resetting master branch\e[0m"
+git reset --hard
+
+echo -e "\e[32m>>>> find/checkout release branch\e[0m"
+# Temporarily disable error catching, as it will cause the script to halt
+# Normally this is not needed, but apparently our chosen branch name for famedly
+# releases causes fatal errors if it does not already exist. I believe it is caused by
+# the '/' in the branch name and how that can be a confusing reference for git to try
+# and sort out.
+# Specifically `famedly-release/vX.Y` comes across as 'fatal' instead of just
+# 'non-existent'. I choose to blame the '/'
+set +e
+
+# Expect this a '0' for exists and a '1' for not-exists. Fatal is a '128'. We apply a
+# NOT so the condition only runs if there was an error
+if ! git switch --quiet "$release_branch_name"; then
+  echo "branch did not exist, creating"
+  # -c creates the branch at our current checkout. Which should be a clean checkout of
+  # the master branch from a moment ago
+  git switch -c "$release_branch_name"
+fi
+# Have to re-enable error catching in case something else breaks
+set -e
+
+# Fetch existing tags on this branch. If this was a new branch from main, ours will be the
+# last tag added. If this was an existing branch we are appending to, ours will still be
+# the last tag added.
+last_tag_on_this_branch=$(git describe --tags --abbrev=0)
+echo "last_tag_on_this_branch $last_tag_on_this_branch"
+
+# Our patch level version, everything after the `_`
+famedly_patch_level=${last_tag_on_this_branch#*_}
+echo "famedly_patch_level ${famedly_patch_level}"
+
+# The major, minor and patch level, so `vX.Y.Z`
+long_ver=${last_tag_on_this_branch%_*}
+echo "long_ver ${long_ver}"
+
+# Just the patch level integer, so `Z`
+patch_level=${long_ver##*.}
+echo "patch_level $patch_level"
+
+# The `vX.Y`
+major_minor=${long_ver%.*}
+echo "major_minor $major_minor"
+
+# and just the `Y`
+minor_ver=${major_minor##*.}
+echo "minor_ver $minor_ver"
+
+
+# First determine if the upstream branch has incremented it's patch level
+upstreams_last_tag=$(git ls-remote --tags upstream "$release_name"* | grep -o 'refs/tags/v[0-9]*\.[0-9]*\.[0-9]*' | sort -rV | head -n1 | cut --delimiter='/' --fields=3)
+echo "upstreams_last_tag $upstreams_last_tag"
+upstream_patch_level=${upstreams_last_tag##*.}
+echo "upstream_patch_level $upstream_patch_level"
+upstream_major_minor=${upstreams_last_tag%.*}
+echo "upstream_major_minor $upstream_major_minor"
+upstream_minor_ver=${upstream_major_minor##.*}
+
+# if upstream patch level has increased, we bump ours too
+if [[ "$expected_minor" != "$minor_ver" ]]; then
+  # must have been a version bump which means this is a new minor version release. In this case, it's likely that it is the same as what upstream's is
+  expected_new_version_tag="${upstreams_last_tag}_1"
+  echo -e "Likely new release minor version $expected_new_version_tag"
+
+elif [[ "$upstream_patch_level" -gt "$patch_level" ]]; then
+  # no minor version bump but there was a patch level version bump, increment our expected patch level
+  expected_new_version_tag="${upstreams_last_tag}_1"
+  echo -e "Likely new release patch version $expected_new_version_tag"
+
+else
+  # no minor and no patch level increase, this is probably just a famedly patch level increase
+  increment_patch_level=$(( famedly_patch_level + 1))
+  expected_new_version_tag="${upstreams_last_tag}_$increment_patch_level"
+  echo -e "Likely new famedly release patch version $expected_new_version_tag"
+
+fi
+
+echo -e "\e[32m>>>> merging master into release branch\e[0m"
+# We let this error if there was a merge conflict that needs to be resolved. It should
+# be an idempotent process of just rerunning the script to continue on. After a
+# `git merge --continue`, of course
+git merge "upstream/release-${release_name}" -m "Famedly ${release_name}"
+
+
+# Retrieve the git log entries themselves and format them to fit well into the
+# markdown format. If there were none, we can skip this whole next section
+git_log_output=$(git log $(git describe --tags --abbrev=0)..master-test --pretty=format:'- %s %C(bold blue)(%an)%Creset\' --no-merges --no-decorate | sort -u -f -k2)
+
+if [[ ! -z $git_log_output ]]; then
+  echo -e "\e[32m>>>> Listing changes from main branch since last release\e[0m"
+  # Find out how many lines down in CHANGES.md that the previous '# Synapse' entry
+  # was. This gives us the line number needed to run the insertion.
+  changelog_line_number=$(grep -n "# Synapse" CHANGES.md | awk 'NR==2 {print $1}' FS=":")
+
+  # Shell expansion is stupid when it comes to escaping things. This is a hack to get the
+  # new line character into a string that sed will actually interpret that does not
+  # interfere with already appended '\' from the git log output. Used at the end to give
+  # us the extra line inside the markdown.
+  NL=$'\n'
+  # Construct the sed command. This is a little messy. sed requires that new lines are a
+  # '\' and not a normal ansi '\n'(and it will insert that for us). We need a basic header
+  # for the entry and an additional new line at the end(the markdown itself does not care
+  # but this makes it easier to find when opening the file directly).
+  sed_command="${changelog_line_number} i\### Famedly additions for ${expected_new_version_tag}\n\n"
+  sed_command+="${git_log_output}${NL}"
+
+  # That should do it. The changes from our repo that occurred since the last upstream
+  # merge(specifically that do not include said merge) should be inserted just above that
+  # previous upstream merge(and just underneath the merge we are in the middle of).
+  # Nice little bonus, they are printed in the console for us to see as the script runs
+  sed -i "${sed_command}" CHANGES.md
+
+  # Amend the last commit(which should be the merge commit itself) to incorporate changes
+  # from the changelog just produced
+  git add CHANGES.md
+  git commit --amend --no-edit --allow-empty
+fi
+#echo "Check any changes or adjustments that need to be made, then"
+#read -n 1 -s -r -p "Press any key to continue merge..."
+
+read -n 1 -p "Press 'p' to push branch to Github, or any other key to skip\n" input_key
+if [[ $input_key == "p" ]]; then
+  echo -e "\e[32m>>>> pushing release branch\e[0m"
+  git push --force -u origin "${release_branch_name}"
+fi
+
+read -n 1 -p "Press 't' to run linting and tests, or any other key to skip\n" input_key
+if [[ $input_key == "t" ]]; then
+  echo -e "\e[32m>>>> running lint and tests...\e[0m"
+  # Make sure there are no weirdities around poetry. Until the deprecation migration
+  # occurs, expect orange things to read here.
+  #poetry check
+  #poetry install --extras all --no-interaction
+  #poetry run ./scripts-dev/lint.sh
+  #poetry run trial -j"${logical_cores}" tests
+  echo -e "\e[32m>>>> Success!\e[0m"
+fi
+
+read -n 1 -p "Press 't' to create tag and push to Github, or any other key to skip\n" input_key
+if [[ $input_key == "t" ]]; then
+  echo -e "\e[32m>>>> updating release tag\e[0m"
+  git tag -f -s -m "${expected_new_version_tag}" "${expected_new_version_tag}"
+#  git push -f origin "${expected_new_version_tag}"
+fi
+echo -e "\e[32m>>>> Finished!\e[0m"