mirror of
https://gitlab.gnome.org/GNOME/glib.git
synced 2024-09-20 01:06:15 +02:00
1378a40734
Rather than fetching the entire repository history, limit the fetch depth to commits in the last 4 weeks. If a branch was branched longer than 4 weeks ago, it seems reasonable to require it to be rebased before it can be reviewed. At the current rate of development, this reduces the bandwidth needed for git pulls in the initial code checks from 73MB to 9MB. Signed-off-by: Philip Withnall <withnall@endlessm.com>
55 lines
2.6 KiB
Bash
Executable File
55 lines
2.6 KiB
Bash
Executable File
#!/bin/bash
|
||
|
||
set -e
|
||
|
||
ancestor_horizon=28 # days (4 weeks)
|
||
|
||
# Wrap everything in a subshell so we can propagate the exit status.
|
||
(
|
||
|
||
# We need to add a new remote for the upstream target branch, since this script
|
||
# could be running in a personal fork of the repository which has out of date
|
||
# branches.
|
||
#
|
||
# Limit the fetch to a certain date horizon to limit the amount of data we get.
|
||
# If the branch was forked from origin/master before this horizon, it should
|
||
# probably be rebased.
|
||
git remote add upstream https://gitlab.gnome.org/GNOME/glib.git
|
||
git fetch --shallow-since="$(date --date="${ancestor_horizon} days ago" +%Y-%m-%d)" upstream
|
||
|
||
# Work out the newest common ancestor between the detached HEAD that this CI job
|
||
# has checked out, and the upstream target branch (which will typically be
|
||
# `upstream/master` or `upstream/glib-2-62`).
|
||
#
|
||
# `${CI_MERGE_REQUEST_TARGET_BRANCH_NAME}` is only defined if we’re running in
|
||
# a merge request pipeline; fall back to `${CI_DEFAULT_BRANCH}` otherwise.
|
||
newest_common_ancestor_sha=$(diff --old-line-format='' --new-line-format='' <(git rev-list --first-parent "upstream/${CI_MERGE_REQUEST_TARGET_BRANCH_NAME:-${CI_DEFAULT_BRANCH}}") <(git rev-list --first-parent HEAD) | head -1)
|
||
if [ -z "${newest_common_ancestor_sha}" ]; then
|
||
echo "Couldn’t find common ancestor with upstream master. This typically"
|
||
echo "happens if you branched from master a long time ago. Please update"
|
||
echo "your clone, rebase, and re-push your branch."
|
||
exit 1
|
||
fi
|
||
|
||
git diff -U0 --no-color "${newest_common_ancestor_sha}" | ./clang-format-diff.py -binary "clang-format-7" -p1
|
||
|
||
)
|
||
exit_status=$?
|
||
|
||
# The style check is not infallible. The clang-format configuration cannot
|
||
# perfectly describe GLib’s coding style: in particular, it cannot align
|
||
# function arguments. The documented coding style for GLib takes priority over
|
||
# clang-format suggestions. Hopefully we can eventually improve clang-format to
|
||
# be configurable enough for our coding style. That’s why this CI check is OK
|
||
# to fail: the idea is that people can look through the output and ignore it if
|
||
# it’s wrong. (That situation can also happen if someone touches pre-existing
|
||
# badly formatted code and it doesn’t make sense to tidy up the wider coding
|
||
# style with the changes they’re making.)
|
||
echo ""
|
||
echo "Note that clang-format output is advisory and cannot always match the GLib coding style, documented at"
|
||
echo " https://gitlab.gnome.org/GNOME/gtk/blob/master/docs/CODING-STYLE"
|
||
echo "Warnings from this tool can be ignored in favour of the documented coding style,"
|
||
echo "or in favour of matching the style of existing surrounding code."
|
||
|
||
exit ${exit_status}
|