镜像站点 · 本页由第三方 GitHub 只读镜像提供,非 GitHub 官方站点,不接受任何登录或凭据输入。前往 github.com
Skip to content

[NEXT-SECURITY-RELEASE] Heads up on upcoming Node.js security release 2025/12/15 #4200

Closed

Description

@marco-ippolito

As per security release workflow, creating issue to give the build team a heads up.

Activity

  1. richardlau commented on Dec 8, 2025

    @richardlau
    Member

    @marco-ippolito From when do you need the CI locked down from?

  2. marco-ippolito commented on Dec 8, 2025

    @marco-ippolito
    MemberAuthor

    I think Wednesday, since are releasing on Monday and its 4 lines

    Given we might have to delay due to a pending fix, I'd say Thursday 11th

  3. richardlau commented on Dec 10, 2025

    @richardlau
    Member

    Okay, I'll the CI lock down from tomorrow.

  4. richardlau commented on Dec 11, 2025

    @richardlau
    Member

    CI is now locked down.

  5. richardlau commented on Dec 17, 2025

    @richardlau
    Member

    The security release has been delayed until 7 January 2026.

    There has been an ask if we can partially relax the controls on the public Jenkins CI in the meantime to allow CI on non-security related PRs.

    Specifically, the ask/proposal was:

    • Allow TSC members to start CI runs.
    • Allow the github-bot to start CI runs.

    To recap, we lock down the CI primarily to avoid any potential leakage of the security patches based on the CI runs.

    For the first point, TSC members currently do not have permission to start CI during a CI lockdown to avoid tying up the CI and give priority to the people preparing the security release. TSC members already have full access to the Hackerone reports and the private repo hosting the proposed patches, so there is no additional leakage risk here.

    For the github-bot, usually it is considered a collaborator (it is a member). Access to the bot account is @nodejs/build-infra, who should already have full access to Jenkins. So I think it should be okay to give the bot permission to start CI runs.

    If the above is done, then for the next few weeks CI can be run:

    • by the request-ci label in GitHub
    • by a TSC member

    Since general read access would still be restricted, it would be up to TSC members to relay any information from the CI runs (i.e. failures).

    @nodejs/build If I don't hear any objections/concerns with the plan I'll go ahead either Thursday or Friday.

  6. Renegade334 commented on Dec 19, 2025

    @Renegade334
    Member

    Just an observer's perspective, but might it be better to keep the request-ci workflow disabled, and have manual runs only?

    I'm just thinking that releasing a couple of dozen PRs into the CI pipeline simultaneously (almost all of which are inevitably going to need resuming for flakes at least once) is going to put a lot of burden on TSC members, particularly given the time of year. If kept manual-only, then any members who wish to volunteer themselves to "take ownership" of a particular PR's CI run can do so, without wasting runner time on jobs that no-one ends up following up and need to be fully re-run at a later date.

  7. richardlau commented on Dec 19, 2025

    @richardlau
    Member

    I've relaxed the controls as per the proposal in #4200 (comment):

    Image

    Just an observer's perspective, but might it be better to keep the request-ci workflow disabled, and have manual runs only?

    The change should allow the bot to be able to start CI runs. I'm going to defer whether to reenable the workflow to other @nodejs/tsc members who are likely to be active over the next few weeks.

  8. joyeecheung commented on Dec 19, 2025

    @joyeecheung
    Member

    Just an observer's perspective, but might it be better to keep the request-ci workflow disabled, and have manual runs only?

    I think the concern is more about read access than write access (or access to start a CI run), because read access can potentially leak information about the vulnerabilities being tested in the same CI infra (that persists information and cache over runs).

    Jenkins doesn't seem to allow write access without read access, so that configuration might not be possible; but also, since the visibility of the the CI results need to be restricted to those who also can access the vulnerabilities, even if that configuration is possible it probably is not too different in terms of workload - even if those who don't have access to the vulnerabilities can restart CI themselves, they still need to ping those who do to see whether the last CI run succeeds to decide whether a re-run is needed or whether a PR is ready to land, or to understand how it fails if the failure is real.

  9. RafaelGSS commented on Jan 13, 2026

    @RafaelGSS
    Member

    Security release is now available.

  10. marco-ippolito commented on Jan 13, 2026

    @marco-ippolito
    MemberAuthor

    Security release is out

  11. richardlau commented on Jan 13, 2026

    @richardlau
    Member

    CI has now been unlocked.

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Metadata

Metadata

Assignees

No one assigned

    Type

    No type

    Projects

    No projects

      Milestone

      No milestone

      Relationships

      None yet

      Development

      No branches or pull requests

      Issue actions