Publish a comment from CircleCI to Github

.circleci/config.yml

@@ -154,6 +154,11 @@ orbs:
               command: |
                   tools/circleci-prepare-log-dir.sh
                   if [ -n "${AWS_SECRET_ACCESS_KEY}" ]; then tools/circleci-upload-to-s3.sh; fi
+          - run:
+              name: Publish a comment to GitHub
+              when: always
+              command: |
+                  tools/circle-publish-github-comment.sh
 
       package:
         parallelism: 1
@@ -265,6 +270,11 @@ orbs:
               command: |
                   tools/circleci-prepare-log-dir.sh
                   if [ -n "${AWS_SECRET_ACCESS_KEY}" ]; then tools/circleci-upload-to-s3.sh; fi
+          - run:
+              name: Publish a comment to GitHub
+              when: always
+              command: |
+                  tools/circle-publish-github-comment.sh
 
       docker_image:
         parallelism: 1

tools/circle-publish-github-comment.sh

@@ -0,0 +1,228 @@
+#!/usr/bin/env bash
+
+source tools/circleci-helpers.sh
+
+set -e
+
+if [ -z "$CIRCLE_BRANCH" ]; then
+    echo "\$CIRCLE_BRANCH is empty. Do nothing"
+    exit 0
+fi
+
+if [ -z "$CIRCLE_PULL_REQUEST" ]; then
+    echo "\$CIRCLE_PULL_REQUEST is empty. Do nothing"
+    exit 0
+fi
+
+if [ "$CIRCLE_PULL_REQUEST" = "false" ]; then
+    echo "Not a pull request. Do nothing"
+    exit 0
+fi
+
+if [ -z "$COMMENTER_GITHUB_TOKEN" ]; then
+    echo "\$COMMENTER_GITHUB_TOKEN is empty. Do nothing"
+    exit 0
+fi
+
+
+# COMMENTER_GITHUB_USER is nickname of a special github user.
+# COMMENTER_GITHUB_TOKEN is token with scope public repo.
+# Token can be obtained here https://github.com/settings/tokens/new
+# Don't use ANY REAL user as a commenter, because GitHub access scopes are too wide.
+echo "Used vars (you can use them to debug locally):"
+echo "export COMMENTER_GITHUB_TOKEN=fillme"
+echo "export COMMENTER_GITHUB_USER=$COMMENTER_GITHUB_USER"
+echo "export PRESET=$PRESET"
+echo "export CIRCLE_SHA1=$CIRCLE_SHA1"
+echo "export CIRCLE_BUILD_URL=$CIRCLE_BUILD_URL"
+echo "export CIRCLE_JOB=$CIRCLE_JOB"
+echo "export CIRCLE_PROJECT_USERNAME=$CIRCLE_PROJECT_USERNAME"
+echo "export CIRCLE_PROJECT_REPONAME=$CIRCLE_PROJECT_REPONAME"
+echo "export CIRCLE_PULL_REQUEST=$CIRCLE_PULL_REQUEST"
+
+PR_NUM="${CIRCLE_PULL_REQUEST##*/}"
+
+REPO_SLUG="$CIRCLE_PROJECT_USERNAME/$CIRCLE_PROJECT_REPONAME"
+
+# IF we only run small tests, the file is missing
+touch /tmp/ct_markdown
+
+PRESET="${PRESET:-default}"
+
+function remove_ct_log_links
+{
+    mv /tmp/ct_markdown /tmp/ct_markdown_original
+    # grep fails if nothing matches
+    { grep -v "Report log" /tmp/ct_markdown_original || echo ""; } > /tmp/ct_markdown
+}
+
+# https://stackoverflow.com/questions/407523/escape-a-string-for-a-sed-replace-pattern
+function replace_string {
+    sed -i "s/$( \
+        echo "$1" | sed -e 's/\([[\/.*]\|\]\)/\\&/g' \
+                  | sed -e 's:\t:\\t:g' \
+    )/$( \
+        echo "$2" | sed -e 's/[\/&]/\\&/g' \
+                  | sed -e 's:\t:\\t:g' \
+    )/g" "$3"
+}
+
+function rewrite_log_links_to_s3
+{
+    local CT_REPORT=big_tests/ct_report
+    local CT_REPORT_ABS=$(./tools/abs_dirpath.sh "$CT_REPORT")
+    local CT_REPORTS=$(ct_reports_dir)
+    local BIG_TESTS_URL="$(direct_s3_url ${CT_REPORTS})/big"
+    cp /tmp/ct_markdown /tmp/ct_markdown_original
+    replace_string "$CT_REPORT_ABS" "$BIG_TESTS_URL" /tmp/ct_markdown
+    # URL escape for s3_reports.html script
+    replace_string "ct_run.test@" "ct_run.test%40" /tmp/ct_markdown
+}
+
+function last_ct_run_name
+{
+    ls -1 -t big_tests/ct_report/ | grep ct_run | head -n1
+}
+
+
+function small_suite_path
+{
+    if [ -d _build/test/logs ]; then
+        cd _build/test/logs
+        ls -t -1 ct_run.mongooseim@localhost.*/lib.mongooseim.logs/run.*/suite.log.html
+        cd ../../..
+    fi
+}
+
+function ct_run_url
+{
+    local CT_REPORTS=$(ct_reports_dir)
+    local BIG_TESTS_URL="$(direct_s3_url ${CT_REPORTS})/big"
+    local RUN_PART=$(echo "$(last_ct_run_name)" | sed "s/@/%40/g")
+    echo "$BIG_TESTS_URL/$RUN_PART/index.html"
+}
+
+function ct_small_url
+{
+    local CT_REPORTS=$(ct_reports_dir)
+    local SMALL_TESTS_URL="$(direct_s3_url ${CT_REPORTS})/small"
+    local SUFFIX=$(small_suite_path)
+    echo "$SMALL_TESTS_URL/$SUFFIX"
+}
+
+function reports_url
+{
+    local CT_REPORTS=$(ct_reports_dir)
+    # Something like that:
+    # http://esl.github.io/circleci-mim-results/s3_reports.html?prefix=PR/3173/60934/pgsql_mnesia.23.0.3-1
+    s3_url "${CT_REPORTS}"
+}
+
+if [ -z "$AWS_SECRET_ACCESS_KEY" ]; then
+    REPORTS_URL_BODY="Reports are not uploaded"$'\n'
+    remove_ct_log_links
+else
+    REPORTS_URL="$(reports_url)"
+    CT_RUN_URL="$(ct_run_url)"
+    SMALL_CT_URL="$(ct_small_url)"
+    REPORTS_BIG_URL_BODY=""
+    REPORTS_SMALL_URL_BODY=""
+    if [ -f big_tests/ct_report/index.html ]; then
+        rewrite_log_links_to_s3
+        REPORTS_BIG_URL_BODY="/ [big]($CT_RUN_URL)"
+    fi
+    if [ -d _build/test/logs ]; then
+        REPORTS_SMALL_URL_BODY=" / [small]($SMALL_CT_URL)"
+    fi
+    REPORTS_URL_BODY="Reports [root](${REPORTS_URL})${REPORTS_BIG_URL_BODY}${REPORTS_SMALL_URL_BODY}"$'\n'
+fi
+
+COUNTERS_FILE=/tmp/ct_stats_vars
+COUNTERS_BODY=""
+if [ -f "$COUNTERS_FILE" ]; then
+    . "$COUNTERS_FILE"
+    COUNTERS_BODY="**OK: $CT_COUNTER_OK** "
+    if [ "$CT_COUNTER_FAILED" != "0" ]; then
+        COUNTERS_BODY="$COUNTERS_BODY/ **Failed: $CT_COUNTER_FAILED** "
+    else
+        COUNTERS_BODY="$COUNTERS_BODY/ Failed: 0 "
+    fi
+    if [ "$CT_COUNTER_USER_SKIPPED" != "0" ]; then
+        COUNTERS_BODY="$COUNTERS_BODY/ **User-skipped: $CT_COUNTER_USER_SKIPPED** "
+    else
+        COUNTERS_BODY="$COUNTERS_BODY/ User-skipped: 0 "
+    fi
+    if [ "$CT_COUNTER_AUTO_SKIPPED" != "0" ]; then
+        COUNTERS_BODY="$COUNTERS_BODY/ **Auto-skipped: $CT_COUNTER_AUTO_SKIPPED** "
+    else
+        COUNTERS_BODY="$COUNTERS_BODY/ Auto-skipped: 0 "
+    fi
+    COUNTERS_BODY="$COUNTERS_BODY"$'\n'
+fi
+
+TRUNCATED_BODY=""
+# Number of truncated failed tests if file exists
+TRUNCATED_FILE="/tmp/ct_markdown_truncated"
+if [ -f "$TRUNCATED_FILE" ]; then
+    TRUNCATED_COUNTER=$(cat "$TRUNCATED_FILE")
+    TRUNCATED_BODY=$'\n'$'\n'"$TRUNCATED_COUNTER errors were truncated"
+fi
+
+# Link to a CircleCI job
+JOB_URL="$CIRCLE_BUILD_URL"
+DESC_BODY="[$CIRCLE_JOB]($JOB_URL) / $PRESET / $CIRCLE_SHA1"$'\n'
+# This file is created by ct_markdown_errors_hook
+ERRORS_BODY="$(cat /tmp/ct_markdown || echo '/tmp/ct_markdown missing')"
+BODY="${DESC_BODY}${REPORTS_URL_BODY}${COUNTERS_BODY}${ERRORS_BODY}${TRUNCATED_BODY}"
+
+function post_new_comment
+{
+# Create a comment GitHub API doc
+# https://developer.github.com/v3/issues/comments/#create-a-comment
+echo "Posting a new comment"
+POST_BODY=$(BODY_ENV="$BODY" jq -n '{body: env.BODY_ENV}')
+curl -o /dev/null -i \
+    -H "Authorization: token $COMMENTER_GITHUB_TOKEN" \
+    -H "Content-Type: application/json" \
+    -X POST -d "$POST_BODY" \
+    https://api.github.com/repos/$REPO_SLUG/issues/$PR_NUM/comments
+}
+
+function append_comment
+{
+# Concat old comment text and some extra text
+# Keep a line separator between them
+# $'\n' is a new line in bash
+#
+# Edit commment GitHub API doc
+# https://developer.github.com/v3/issues/comments/#edit-a-comment
+COMMENT_ID="$1"
+echo "Patch comment $COMMENT_ID"
+BODY_FROM_GH="$(cat /tmp/gh_comment | jq -r .body)"
+BODY="${BODY_FROM_GH}"$'\n'$'\n'"---"$'\n'$'\n'"${BODY}"
+PATCH_BODY=$(BODY_ENV="${BODY}" jq -n '{body: env.BODY_ENV}')
+curl -o /dev/null -i \
+    -H "Authorization: token $COMMENTER_GITHUB_TOKEN" \
+    -H "Content-Type: application/json" \
+    -X PATCH -d "$PATCH_BODY" \
+    https://api.github.com/repos/$REPO_SLUG/issues/comments/$COMMENT_ID
+}
+
+# List comments
+# https://developer.github.com/v3/issues/comments/#list-comments-on-an-issue
+curl -s -S -o /tmp/gh_comments -L \
+    -H "Authorization: token $COMMENTER_GITHUB_TOKEN" \
+    -H "Content-Type: application/json" \
+    https://api.github.com/repos/$REPO_SLUG/issues/$PR_NUM/comments
+
+# Filter out all comments for a particular user
+# Then filter out all comments that have a git commit rev in the body text
+# Then take first comment (we don't expect more comments anyway)
+cat /tmp/gh_comments | jq "map(select(.user.login == \"$COMMENTER_GITHUB_USER\")) | map(select(.body | contains(\"$CIRCLE_SHA1\")))[0]" > /tmp/gh_comment
+COMMENT_ID=$(cat /tmp/gh_comment | jq .id)
+
+if [ "$COMMENT_ID" = "null" ]; then
+    post_new_comment
+else
+    append_comment "$COMMENT_ID"
+fi