Conversation

Notices

1. Embed this notice
   Chris Vest (chrisvest@mastodon.social)'s status on Friday, 13-Jun-2025 23:48:34 JST Chris Vest

   We got the conclusion to jemalloc. Jason Evans has written a post mortem: https://jasone.github.io/2025/06/12/jemalloc-postmortem/

   • Haelwenn /элвэн/ :triskell: repeated this.
   • Embed this notice
     Poul-Henning Kamp (bsdphk@fosstodon.org)'s status on Friday, 13-Jun-2025 23:48:33 JST Poul-Henning Kamp

     in reply to

     @chrisvest

     I guess I have to send him a laminated membership card to the Retired FreeBSD malloc Maintainers Association now ?

     https://phk.freebsd.dk/sagas/phkmalloc/

     Haelwenn /элвэн/ :triskell: likes this.
   • Embed this notice
     Poul-Henning Kamp (bsdphk@fosstodon.org)'s status on Saturday, 14-Jun-2025 04:40:58 JST Poul-Henning Kamp

     in reply to

     @chrisvest

     Yes, not a bad one either.

     But "alternative" malloc implementations need people to actively decide to use them, which is reactive to some specific problem.

     phkmalloc, by being OS default, meant the 10K+ ports collection got exposed to it.

     I remember only one or two "Bah!, FreeBSD is broken watch me not care!" upstream reactions, everybody else went "Ohh, I guess everybody but FreeBSD are unsafe!" and fixed the reported bugs.

     Some even said thanks :-)

     Haelwenn /элвэн/ :triskell: likes this.
   • Embed this notice
     Chris Vest (chrisvest@mastodon.social)'s status on Saturday, 14-Jun-2025 04:40:59 JST Chris Vest

     in reply to

     • Poul-Henning Kamp

     @bsdphk Nice read. It reminds me Bonwick's Slab allocator also placed debugging and error detection front and center.