[Looking for Charlie's main web site? or all posts?]

Announcing Java updates of Oct 20 2026 - thoughts and resources

It's that time again: there are new Oracle JVM updates released today (Oct 20, 2026) for the current long-term support (LTS) releases of Oracle Java 25, 21, 17, 11, and 8, as well as the newest short-term release, 27. (Yep, there are for now 5 current Oracle Java "LTS" releases, for historical reasons.)
Update: Oct 21, 2026: Within hours of the release of that update, Oracle has released a critical security point-release update. The new downloads (if you obtain them from that point forward) now add a .0.2 suffix to all the numbers below, such as 21.0.13.0.2, 17.0.21.0.2 and so on. For more on the vuln that triggered that, which has a 9.1 cvss score (and is "remotely exploitable without authentication"), see the Oracle critical alert advisory notice on the matter.

Update: Oct 24, 2026: Just a couple of days later, Oracle has released yet another a point-release update--related very much to that previous one, this time removing the set of impacted root CA certificates related to eMudhra. The new downloads (if you obtain them from Oracle that date forward) now add a .0.3 suffix to all the numbers below, such as 21.0.13.0.3, 17.0.21.0.3 and so on. For more on that, which again has a 9.1 cvss score ("remotely exploitable without authentication"), see the Oracle critical alert advisory notice on the matter, and you can see this and the last one in a list of all such jvm security updates. And for more from Oracle on the eMudhra certs--including options to remediate manually instead of getting this .0.3 update, see this notice.

Now back to the rest of the post, which otherwise still stands.

TLDR: The new updates are 27.0.1, 25.0.5, 21.0.13, 17.0.21, 11.0.33, and 1.8.0_511 (aka 8u511) respectively. From a compatibility perspective, the update seems a pretty modest one, without seeming breaking changes (though my opinion is based solely on my read of the update release notes, on this first day).

More on the updates below, including links to more info on each of them including what changed, bug fixes, and the security fixes each version contains.

[....Continue Reading....]

Comments
Oracle came out with a critical security update as a point release just hours after this release, so the new version numbers (when downloaded from Oracle) currently end with a suffix of .0.2, as in 21.0.5.0.2 and so on. For more, see an update I added to the top of this blog post.
Asked Adobe CF Support about the lack of JVM on the downloads page and if we are still covered if we download the JVM from "non adobe" site, here is reply:

Download the latest from here: https://cfdownload.adobe.com/pub/adobe/coldfusion/java/java17/java1702102/jdk-17.0.21.0.2_windows-x64_bin.exe

The license would be covered if they are downloaded from our site. Refer to this article from our Product Manager: https://coldfusion.adobe.com/2023/11/oracle-java-support-for-adobe-coldfusion-2021-2023/
# Posted By Joseph C | 10/22/26 3:48 PM
Just a note, applied an update tonight ( 8u511 ) on some real old servers and had to add the following to the jvm.config so DB connections and for Fusion Reactor to be able to ping their licensing service. Basically, was required for any HTTPS connection that required newer certificate and key features.

-Dcoldfusion.disablejsafe=true

For the version I had to apply this on, this was the reason as per Claude and thanks to a fellow developer that jumped on to help.

ColdFusion ships RSA BSAFE (JSafe/JCE) as a security provider, and it sits ahead of the JDK’s default providers. On a newer 8u build, the older bundled JSafe doesn’t handle some newer certificate and key features, such as modern Let’s Encrypt chains (ECDSA/ISRG X2 and newer signature handling). That produces handshake or path-validation failures even though the roots are in cacerts. Disabling it makes CF fall back to the JDK’s own SunJSSE/SunEC providers, which is why the trust store looked fine but connections still failed.
As yet another update (after the initial release of this Oct '26 Java cpu release), Oracle has released yesterday yet another point release update, this time with the version number ending in a .0.3 suffix. See my brief discussion and links to more info in my "update" notices at the top of this post. Both the updates relate to eMudhra certs--but note that that the issue IS NOT about whether YOU intentionally use them or not but how hackers could misuse them before the update.

Note also that I offer there a link to an Oracle technote discussing manually removing the affected immediate and root certs, for those who may not want to update the jvm yet again just for the 0.3 version. More at https://ops.java/emudhra .
Joseph, yes, Adobe did make this new Oct '26 Java CPU version available a couple days after my post. I've now updated my wording above to reflect that. Better still, they did also implement that further-updated 0.2 version that I discuss above.

They have NOT yet added the further-updated 0.3 version, and I won't be surprised if they don't. Folks will therefore want to note how I share in my updates above how one can manually remediate what's addressed in that 0.3 update. Or they could obtain it from Oracle, as I also discuss above.

But finally you quote Adobe support saying, "The license would be covered if they are downloaded from our site." I understand that the implication here is that you must. I'll say again: the binary files are identical, regardless of where downloaded (assuming the exact same point release, of course). And the license is IN that file. I've never seen any formal verbiage (about WHERE downloaded from) beyond such vague assertions.

In fact even that blog link they offered you does NOT assert that we MUST get it from them. But FWIW, IANAL.
Giancarlo, thanks for your comment. That said, cf hasn't run on Java 8 for over a decade. Perhaps you're using the CF Api Manager, which curiously Adobe still has left running on Java 8--and thankfully Oracle is indeed still updating that.

And perhaps there may well be some instance where someone running cf on later Java versions can still benefit from that info and the jvm arg. That said, I do hope that anyone running cf is on the versions which Adobe still updates, which are currently cf2025 and cf2023, and which run Java 21 and 17, respectively.
Will add to the discussion about downloading from Adobe. As I type this Monday, October 26 at 1040AM EDT, Adobe's download page still doesn't have .3 available.

Frustrating if Adobe is saying "if downloaded from our site" implies an issue, yet they don't appear to have a good mechanism to update the downloads page promptly.
# Posted By BrianM | 10/26/26 9:42 AM
But again, there's zero binary difference between what's offered on Adobe's vs Oracle's site. And the license we agree to is in the downloaded file--which is identical therefore in the downloads from either site.

Like I also said, I won't be surprised if they don't get to updating that. As for why it's so challenging for them, I don't know or understand. I've long noted how frustrating it is to get even just the quarterly files posted there--when the dates are announced literally a year in advance!

My take on things is that it's better to light one candle then to curse the darkness.
Copyright ©2026 Charlie Arehart
Carehart Logo
BlogCFC was created by Raymond Camden. This blog is running version 5.005.
(Want to validate the HTML in this page?)

Managed Hosting Services provided by
xByte cloud Hosting