Jeremy Kerr

open source software developer / hardware bringup / whadjuk nyungar boodjar

Jeremy Kerr boosted:
2025-12-14

@NanoRaptor cycling by the river and a crow landed on the bike path in front of me, jumped in my way when i tried to go around so i stopped.

it jumped up and down until i dismounted and followed it to a nearby road where some of it's friends were squawking, all clustered around an endangered species of tortoise that couldn't mount the kerb to get to the river.

picked up the tortoise and walked it to the river, it swam off and the crows happily cawed and then all flew off.

Jeremy Kerrjk@aus.social
2025-11-28

@ricci

#!/usr/bin/env pyþon3

Jeremy Kerrjk@aus.social
2025-11-12

@andyprice may I see it?

Jeremy Kerr boosted:
Andrew Jefferyarj@social.tchncs.de
2025-11-11

Upstream #deku now supports the bit-precise attributes without requiring the alloc feature 🎉

Jeremy Kerrjk@aus.social
2025-11-04
Graph of PostgreSQL connections to patchwork database, showing a steep drop in idle connections at midnight UTC Tuesday
Jeremy Kerrjk@aus.social
2025-11-04

... so, uh, if you're working at TI and your patchwork downloader has stopped working, maybe get in touch?

Jeremy Kerrjk@aus.social
2025-11-04

patchwork.ozlabs.org has been slow on response times lately, and I have recently had a few emails from project maintainers asking why. I had managed to get some time to look into it over the last few days.

CPU load seems fine; a fair amount of IO, but nothing too severe. Typically around 15 req/sec. No specific system issues, but a lot of waiting on either wsgi handlers, or postgres transactions (the former likely due to the latter).

Looking at the request logs, a few culprits: a single IP (in Texas Instruments' allocation) repeatedly downloading all u-boot patches over the last four years, at around 300kB per request. Less obvious, but a whole /22 of Alibaba Singapore iterating through the /api URL space, using a generic Safari on Windows user agent.

Both have been blocked, responses seem much more snappy now.

Jeremy Kerrjk@aus.social
2025-10-23

@abstractcode @jpm location *and time*!

Jeremy Kerrjk@aus.social
2025-10-14

@hailey announcing cornmon, the corn monitoring interface

Jeremy Kerrjk@aus.social
2025-09-26

@airlied wait, we're allowed to do "regular programming"?

Jeremy Kerr boosted:
Joel Michaeljpm@aus.social
2025-09-26

@whatifbot that’s just a cat

Jeremy Kerrjk@aus.social
2025-09-26

RE: mastodon.social/@rustfoundatio

Crabtacular! Code Construct has just signed up to support the Rust Foundation.

Jeremy Kerr boosted:
enderman0125 :neofox_flag_nb:mudkip@wetdry.world
2025-09-22
a captcha that asks the user to select all squares with socs that can boot mainline linux, below there are various squares with SoCs in them
Jeremy Kerrjk@aus.social
2025-09-04

@chat8ballbot @emmadavidson isn't this exactly how we got Canberra?

Jeremy Kerr boosted:
2025-08-31

This is what pull requests used to look like.

(The last sentence is particularly remarkable!)

APOLLO SPACECRAFT SOFTWARE CONFIGURATION CONTROL BOARD PROGRAM CHANGE REQUEST
Number: 848

1.1 Originator: D. Eyles, Date: 7/22/69, Organization: MIT/IL
1.2 Approval: George W. Cherry, Date: 7/23/69
1.3 Effectivity: Luminary 1B
1.4 Title of Change: Prevent RR ECDUs from Stealing LGC Memory Cycles
1.5 Reason(s) for Change: See attached Data Amplification Sheet.
1.6 Description of Change: See attached Data Amplification Sheet.

3.0 MIT VISIBILITY IMPACT EVALUATION:
3.1 Schedule Impact: 0
3.2 Storage Impact: ~8 words
3.3 MIT Coordinator: George W. Cherry
3.4 Remarks: Approved by T. Price and T. Gibson on 7-22-69

(and many blank fields)APOLLO SPACECRAFT SOFTWARE CONFIGURATION CONTROL BOARD
DATA AMPLIFICATION SHEET

PROGRAM CHANGE REQUEST #848 8
PREPARED BY: D. EYLES 
DATE: 7/22/69
ORGANIZATION: MIT/IL

CONTINUATION SECTION (REFER TO BLOCK NUMBER AND TITLE
ON PROGRAM CHANGE REQUEST FORM)
1.5 Reason(s) for Change
Because when the Rendezvous Radar Select Switch is not in the LGC position, the RR resolvers are being excited by an 800 cps signal which is not necessarily in phase with the 800 cps reference of the RR CDUs, these CDUs are likely to go to their maximum rate mode (6400 pps each). This behavior snatches a total of 12800 memory cycles every second, or 15% of LGC time, and can result in 1201 and 1202 software restarts.

1.6 Description of Change
Monitor the RR Power On/ Auto bit (bit 2 of Channel 33) and set the Zero RR CDU
bit (bit 1 of Channel 12) when this bit shows that either the RR circuit breakers are
pulled or the RR switch is in AUTO TRACK or SLEW.

REMARKS
It was the misbehavior in question that jeopardized the lunar landing of July 20, 1969.
Jeremy Kerr boosted:
2025-08-30

How you know you found the good shit:tm:

when the IEEE paper starts invoking faux 3D sigils made of logic symbols

A screenshot of a quadrature CDR circuit that is made up of DFFs around the outside in a ring, all feeding XOR gates, which ghen feed differential amplifiers in the middle
Jeremy Kerrjk@aus.social
2025-07-31

@Kels_316 "this email could have been an email"

Jeremy Kerrjk@aus.social
2025-07-31

@ben_hr @liamvhogan ok, done. Now do I put this "Files" folder in my "Folders" folder? 🤔

Jeremy Kerrjk@aus.social
2025-07-29

@zev @arj eep! fixed, thanks!

Jeremy Kerr boosted:
Andrew Jefferyarj@social.tchncs.de
2025-07-29

Recently we've been putting together a suite of rust crates for #MCTP and related protocols such as PLDM and NVMe-MI, for Linux and embedded environments
codeconstruct.com.au/docs/mctp

Client Info

Server: https://mastodon.social
Version: 2025.07
Repository: https://github.com/cyevgeniy/lmst