drop el-7, debian-buster, amazonlinux-2, CircleCI #14990
Conversation
Habbie
force-pushed
the
drop-el7-debianbuster
branch
from
50272e0 to
70a9a45
Compare
this claim was based on https://endoflife.date/amazon-linux which says 30 Jun 2025 (update submitted). However, https://aws.amazon.com/amazon-linux-2/faqs/ says 30 Jun 2026. That said, I don't think Amazon Linux 2 is reason enough to hang on to all that legacy. I'm not even sure we have any users on it. |
Pull Request Test Coverage Report for Build 12432070620Details
💛 - Coveralls |
|
I cannot offer a strong review but a weak review it eye balls good to me. There are a few spec file comments that refer to EL7 still (see the systemd runtime dir bits) but thats minor. |
ah yes. I missed at least one comment there, thanks! |
| @@ -16,10 +16,8 @@ on: | |||
| type: string | |||
| # please remember to update build-packages.yml as well | |||
| default: >- | |||
| el-7 | |||
| el-8 | |||
There was a problem hiding this comment.
should this list be in sync with the list in builder.yml?
There was a problem hiding this comment.
no, in fact these days perhaps it should explicitly not overlap with it at all I think. builder-dispatch should contain a good default set for shipping packages (even if we don't builder-dispatch for releases any more). builder is older than all the more specific ones, and was for daily testing of distros we -don't- autobuild for in any other way.
builder-dispatch should be in sync with builder-packages (or ideally the list would live in one of them, but that's tricky). builder (which we should maybe give a useful suffix to explain what it does) should then cover distros we do not, or do not yet, ship for
| ./install_quiche.sh; \ | ||
| fi | ||
| RUN cd /pdns/builder-support/helpers/ && ./install_rust.sh; \ | ||
| yum install -y git cmake clang; \ |
There was a problem hiding this comment.
ISTM all these commands want && chaining for resilience, but this is not a new problem
There was a problem hiding this comment.
agreed. Would make sense as a separate commit if we're touching all this anyway.
| RUN touch /var/lib/rpm/* && if $(grep -q 'release 7' /etc/redhat-release); then \ | ||
| scl enable devtoolset-11 -- builder/helpers/build-specs.sh builder-support/specs/dnsdist.spec; \ | ||
| elif $(grep -q 'release 8' /etc/redhat-release); then \ | ||
| RUN touch /var/lib/rpm/* && if $(grep -q 'release 8' /etc/redhat-release); then \ |
There was a problem hiding this comment.
in theory, in this context grep ... should be equivalent to $(grep...)
| @@ -104,9 +91,6 @@ EOF | |||
| # The EL7 and 8 systemd actually supports %t, but its version number is older than that, so we do use seperate runtime dirs, but don't rely on RUNTIME_DIRECTORY | |||
There was a problem hiding this comment.
yes! I think that's what ph1 spotted too
There was a problem hiding this comment.
ah yes! the same comment is copied into a second file too
| @@ -144,35 +123,27 @@ exit 0 | |||
| %if 0%{?suse_version} | |||
| %service_add_post %{name}.service | |||
| %endif | |||
| %if 0%{?rhel} >= 7 | |||
There was a problem hiding this comment.
with the suse context above, this looks like this block might have only been in place for rhel?
Short description
el-7 and debian buster are EOL already. Amazonlinux-2 goes EOL in six months and we never shipped those packages.
We stopped using CircleCI years ago.
To be done (separate PR perhaps): the el7 devtoolset check in misc-dailies should be migrated for el8.
This is entirely untested and needs a strong review because of all the "if el7" things I took out from various files in various syntaxes.
Checklist
I have: