Automate release steps with the Buildkite release triggers

[Sebastian Wilzbach]

Now that we have an easy to edit buildkite pipeline, we could add a few more automations to an automated release trigger (i.e. https://buildkite.com/docs/pipelines/trigger-step).

Ideas:

• docker build
• update the DMD version in the wiki (https://wiki.dlang.org/Template:Latest_DMD_Version_Raw)
• update the dlang.org changelog (at least make the PR)
• send PRs to merge stable back into master
• purge the changelog entries

Martin's release script is here: https://gist.github.com/MartinNowak/a471fe7ddbfeef205cdf04c93a94c6d0

So in short automating bits of it.

Update the wiki

As an example, here's how to update the wiki with a USERNAME + USERPASS.

```bash set -eux -o pipefail

USERNAME="username"
USERPASS="password"
PAGE="Template:Latest_DMD_Version_Raw"
WIKIAPI="https://wiki.dlang.org/api.php"
cookie_jar="/opt/dlang/wikicj"
#Will store file in wikifile

echo "UTF8 check: ☠"
#################login
echo "Logging into $WIKIAPI as $USERNAME..."

text=$(curl --fail --retry 3 http://downloads.dlang.org/releases/LATEST)

###############
#Login part 1
#printf "%s" "Logging in (1/2)..."
echo "Get login token..."
CR=$(curl -S
--location
--retry 2
--retry-delay 5
--cookie $cookie_jar
--cookie-jar $cookie_jar
--user-agent "Curl Shell Script"
--keepalive-time 60
--header "Accept-Language: en-us"
--header "Connection: keep-alive"
--compressed
--request "GET" "${WIKIAPI}?action=query&meta=tokens&type=login&format=json")

echo "$CR" | jq .

rm -f login.json
echo "$CR" > login.json
TOKEN=$(jq --raw-output '.query.tokens.logintoken' login.json)
TOKEN="${TOKEN//"/}" #replace double quote by nothing

#Remove carriage return!
printf "%s" "$TOKEN" > token.txt
TOKEN=$(cat token.txt | sed 's/\r$//')

if [ "$TOKEN" == "null" ]; then
echo "Getting a login token failed."
exit
else
echo "Login token is $TOKEN"
echo "-----"
fi

###############
#Login part 2
echo "Logging in..."
CR=$(curl -S
--location
--cookie $cookie_jar
--cookie-jar $cookie_jar
--user-agent "Curl Shell Script"
--keepalive-time 60
--header "Accept-Language: en-us"
--header "Connection: keep-alive"
--compressed
--data-urlencode "username=${USERNAME}"
--data-urlencode "password=${USERPASS}"
--data-urlencode "rememberMe=1"
--data-urlencode "logintoken=${TOKEN}"
--data-urlencode "loginreturnurl=http://en.wikipedia.org"
--request "POST" "${WIKIAPI}?action=clientlogin&format=json")

echo "$CR" | jq .

STATUS=$(echo $CR | jq '.clientlogin.status')
if [[ $STATUS == "PASS" ]]; then
echo "Successfully logged in as $USERNAME, STATUS is $STATUS."
echo "-----"
else
echo "Unable to login, is logintoken ${TOKEN} correct?"
exit
fi

###############
#Get edit token
echo "Fetching edit token..."
CR=$(curl -S
--location
--cookie $cookie_jar
--cookie-jar $cookie_jar
--user-agent "Curl Shell Script"
--keepalive-time 60
--header "Accept-Language: en-us"
--header "Connection: keep-alive"
--compressed
--request "POST" "${WIKIAPI}?action=query&meta=tokens&format=json")

echo "$CR" | jq .
echo "$CR" > edittoken.json
EDITTOKEN=$(jq --raw-output '.query.tokens.csrftoken' edittoken.json)
rm edittoken.json

EDITTOKEN="${EDITTOKEN//"/}" #replace double quote by nothing

#Remove carriage return!
printf "%s" "$EDITTOKEN" > edittoken.txt
EDITTOKEN=$(cat edittoken.txt | sed 's/\r$//')

if [[ $EDITTOKEN == "+\" ]]; then
echo "Edit token is: $EDITTOKEN"
else
echo "Edit token not set."
exit
fi

###############
#Make a test edit
CR=$(curl -S
--location
--cookie $cookie_jar
--cookie-jar $cookie_jar
--user-agent "Curl Shell Script"
--keepalive-time 60
--header "Accept-Language: en-us"
--header "Connection: keep-alive"
--compressed
--data-urlencode "title=${PAGE}"
--data-urlencode "text=${text}"
--data-urlencode "summary=auto bump"
--data-urlencode "token=${EDITTOKEN}"
--request "POST" "${WIKIAPI}?action=edit&format=json")

echo "$CR" | jq .

</details>

[dawg]

Well the script is automation already and it's unclear whether moving it into a public CI pipeline will work reliably.
It seems more worthwhile to focus on fixing other stuff or improve our CI mess rather than to replace a working system (that's only used monthly) with a new one. Your wiki update script gives a hint at the lingering fragility.