Re: jemalloc 5.3.0 upgrade

> On Mar 4, 2026, at 1:04 AM, dikshie <dikshie@gmail.com> wrote:

…

> Now,  Facebook has unarchived the original jemalloc.
> 
> I just read the announcement in facebook's blog:
> https://engineering.fb.com/2026/03/02/data-infrastructure/investing-in-infrastructure-metas-renewed-commitment-to-jemalloc/

This should be interesting, given that Facebook/Meta has never really gotten itself together around stable release versioning: their whole philosophy is “release from HEAD”, much like Google (at least Google has made concessions with some of the projects it sponsors, like GoogleTest, though). Also, they do not have any interest in maintaining things beyond Linux or platforms that they ship their apps on (AOSP, iOS, Windows, etc), so I don’t foresee an easy path to maintainership on FreeBSD. Finally, Meta’s build process is tightly tied to Buck (a mutant of Bazel), so it’ll be interesting to see what happens in that regard as well. Moving to cmake would be positive, but Buck? Having to run Java/Nailgun to build a trivial library would be overkill (like using burning down a spider to kill a house).

I’d love to be proven wrong, but if past behavior is an accurate predictor for future behavior, I’m not going to hold my breath.

Kind of like Microsoft with the GitHub acquisition and all of the CoPilot enshittification that has ensued (not just in GitHub, but with all Microsoft products).

Thanks,
-Enji

1. Here’s a fun example: https://github.com/facebook/folly/releases . All tags are created automated by a bot, there’re no release notes for each release, and it’s unclear what kind of testing/qualification has been done for each release.