Cherry pick changes for 2.8.5 release #13269
Conversation
mockbin.com redirects to insomnia official site and triggers a security policy, which makes tests fail.
Build 2.8 on Bazel KAG-2835
Several of these tests contained the following assertion after generating a certificate with the `kong hybrid gen_cert` command: ```lua assert(crt:get_not_before() >= ngx.time()) ``` This produces failures every now and again when the clock has advanced _just_ enough for ngx.time() to return `crt:get_not_before() + 1`. To fix this, we record the time _before_ generating the cert and validate against the stored timestamp. (cherry picked from commit b7a8361)
Please check the contained README.md. (cherry picked from commit f135c70) Co-authored-by: Zachary Hu <[email protected]>
… the changelog directory
* chore(test): remove prefix directory when stop_kong called (#12691) If the prefix is not cleaned up when stop_kong is called, it could impact subsequent tests, especially when later tests start Kong by a shell command, the Kong instance might be started up with the default `servroot` prefix. KAG-3808 (cherry picked from commit 3dd5bdb) * remove prefix directory for 2.8
…mit (#12693) (#12711) Replace `${{ secrets.GHA_COMMENT_TOKEN }}` with `${{ secrets.GITHUB_TOKEN }}`. The `${{ secrets.GHA_COMMENT_TOKEN }}` needs to be manually rotated, replacing it by `${{ secrets.GITHUB_TOKEN }}`, which is generated by each run of the workflow, so we don't need to rotate token anymore.
FTI-5842 (cherry picked from commit ed0b96d) Co-authored-by: Zachary Hu <[email protected]>
github-actions
bot
added
the
chore
cherry-pick from Kong/kong-ee#9042 KAG-4775 (cherry picked from commit 8e86dba) Co-authored-by: Niklaus Schen <[email protected]>
* Create a clean version for 2.8 CE. * Add back changelog. --------- Co-authored-by: Zhongwei Yao <[email protected]>
| @@ -0,0 +1,62 @@ | |||
| From 521b367567dc5d91d7f9ae29c257998953e24e53 Mon Sep 17 00:00:00 2001 | |||
There was a problem hiding this comment.
Seems like we accidentally introduced those patches
There was a problem hiding this comment.
Yes, besides that, there are two patches related to M1 that I think can also be removed.
build/openresty/patches/lua-resty-core-0.1.22_03-make-resty.core.shdict-compatible-with-m1.patch
build/openresty/patches/lua-resty-core-0.1.22_04-make-resty.core.response-compatible-with-m1.patch
There was a problem hiding this comment.
2.8.4 (also 2.8.3) were built using kbt 4.25.5 and these patch files were in https://github.com/Kong/kong-build-tools/tree/4.25.5/openresty-patches/patches/1.19.9.1 . During the bazel migration we've migrated them into the kong repository, so they're not new patches.
I've diffed this PR's patch files and the kbt 4.25.5 patches and the only differences are hash fix plus one build config change. The build config change is needed by bazel build:
| @@ -10,3 +10,10 @@ LIBYAML_VERSION=0.2.5 | |||
| KONG_GO_PLUGINSERVER_VERSION=v0.6.1 | |||
| KONG_BUILD_TOOLS_VERSION=4.40.1 | |||
| KONG_NGINX_MODULE_BRANCH=0.2.1 | |||
There was a problem hiding this comment.
It seems pre-bazel, on the 2.8.4 tag, we were:
KONG_PACKAGE_NAME=kong
KONG_CONFLICTS=kong-enterprise-edition
KONG_LICENSE="ASL 2.0"
RESTY_VERSION=1.19.9.1
RESTY_LUAROCKS_VERSION=3.8.0
RESTY_OPENSSL_VERSION=1.1.1o
RESTY_PCRE_VERSION=8.45
LIBYAML_VERSION=0.2.5
KONG_GO_PLUGINSERVER_VERSION=v0.6.1
KONG_BUILD_TOOLS_VERSION=4.40.1
KONG_NGINX_MODULE_BRANCH=0.2.1
There was a problem hiding this comment.
Unless we can't avoid it, if possible, we probably don't want to change any of these versions.
There was a problem hiding this comment.
We downgrade these dependencies' versions to try to keep them consistent with version 2.8.4.
#13282
With this patch, CI will notify a Kong Inc internal slack channel on every PR that performs a schema change.
It seems that if the do not merge label job is skipped then the second job doesn't run either: https://github.com/Kong/kong/actions/runs/4307151445/jobs/7511859202 This change splits the job into two and narrows down the events on which these jobs are triggered since the only meaninful input are the labels on the PR.
This is a bad practice which could cause merge conflicts and is against our backport policy.
It seems that Github Actions is not running these jobs even once even though the PRs are labelled at least once. This patch runs these jobs on other related PR activity.
rebase and merge
Summary
Checklist
changelog/unreleased/kongorskip-changeloglabel added on PR if changelog is unnecessary. README.mdIssue reference
KAG-4786
Fix #[issue number]