Wsrtjtyk

I want to implement Gitlab custom hooks for only one repository using Linux Bash scripts for the following:

1. I want to enforce the developers to mention the JIRA issue number with that issue status (ex: fixed, In progress) in their commit message, if the commit message doesn't have the issue number, the push should be rejected automatically.


2. If the commit message has the JIRA issue ID with the status, the developers should be allowed to push the change.

I am confused between what is the best Gitlab hook I should pick and how do I use JIRA issue in the commit message.

As it happens I have a script that does just that. Mine also checks for compliance with a few other commit style issues but you can customise it to your requirements.

#!/bin/bash

#

# pre-receive hook for Commit Check

#

# Adapted from

#  http://blog.hgomez.net/2015/03/02/Gitlab-custom-hooks-Bash-Way.html



REGEX='^((([A-Z]+-[0-9]+|maint|docs)) [A-Z]|Version [0-9]+).*[^.]$'



check_single_commit()

{

  FIRST_LINE=$(echo "$COMMIT_MESSAGE" | sed -n 1p)

  SECOND_LINE=$(echo "$COMMIT_MESSAGE" | sed -n 2p)



  # Warn if line 1 > 50 chars

  WIDTH=$(echo $FIRST_LINE | wc -c)

  [ $WIDTH -gt 50 ] && [ $WIDTH -le 70 ] && printf "Warning: first line of commit exceeds 50 charactersnn"



  # Fail if there is no Jira, docs or maint ref.

  echo $FIRST_LINE | egrep -q "$REGEX"

  COMMIT_CHECK_STATUS=$?

  if [ $COMMIT_CHECK_STATUS -gt 0 ]; then

    echo "Fail: No Jira ref found in commit message"

    echo "Expected: $REGEX"

    echo "Actual:"

    echo "$COMMIT_MESSAGE" |sed -e 's/^/  /'

    echo ""

  fi



  # Fail if second line is not blank.

  if [ ! -z "$SECOND_LINE" ] && [ $COMMIT_CHECK_STATUS -eq 0 ]; then

    [ -z "$SECOND_LINE" ] ; COMMIT_CHECK_STATUS=$?

    echo "Second line of commit is not blank, but is '$SECOND_LINE'"

    echo ""

  fi



  # Fail if line 1 > 70 chars

  if [ $WIDTH -gt 70 ] && [ $COMMIT_CHECK_STATUS -eq 0 ]; then

    [ $WIDTH -le 70 ] ; COMMIT_CHECK_STATUS=$?

    echo "Fail: first line of commit exceeds 70 characters, please use the commit body"

    echo ""

  fi

}



check_all_commits()

{

  REVISIONS=$(git rev-list master..$NEW_VALUE)

  IFS='n' read -ra LIST_OF_REVISIONS <<< "$REVISIONS"



  for rid in "${!LIST_OF_REVISIONS[@]}"; do

    REVISION=${LIST_OF_REVISIONS[rid]}

    COMMIT_MESSAGE=$(git cat-file commit $REVISION | sed '1,/^$/d')

    check_single_commit



    if [ $COMMIT_CHECK_STATUS -ne 0 ]; then

      echo "Commit validation failed for commit $REVISION" >&2

      echo ""

      echo "Please read: https://chris.beams.io/posts/git-commit/"

      echo ""

      exit 1

    fi

  done

}



# Get custom commit message format



# After a push occurs and before any refs are updated on the remote repository,

# the git-receive-pack process invokes the pre-receive hook script with the

# standard input of one line per ref to be updated:

#

# <old-value> SP <new-value> SP <ref-name> LF

#

# This string represents these arguments:

#

# <old-value> Old object name stored in the ref. When you create a new ref,

# this equals 40 zeroes.

# <new-value> New object name to be stored in the ref. When you delete a ref,

# this equals 40 zeroes.

# <ref-name>  The full name of the ref.



while read OLD_VALUE NEW_VALUE REFNAME ; do

  [ $NEW_VALUE == 0000000000000000000000000000000000000000 ] && exit 0

  check_all_commits

done



exit 0



# vim:set ft=sh