In other news, I poked The Powers That Be to update `pip-tools` so `pip-compile` doesn't explode because of the PEP 517 issue, another piece of breakage I should never have to deal with. Tracking down the source and solution of this problem was not fun.
FYI for anyone getting spurious build failures on recent Python projectsv
Apparently stripping out all license related fields from pyproject.toml and adding `project.license = 'LicenseRef-Other'` is the only way to get our projects to build without error.
This has been a nagging problem for a few months and broke our ability to create release packages. It's amazing how simple the user notification could have been to resolve this self-inflicted problem without torching the entire Python user base's workflow but language and utility developer pique is the primary consideration in this ecosystem. There will be a new gratuitous breaking change to replace this soon enough.
Anyway, hope this resolves someone's unexpected build failures.
In Kent, in England, there’s a town called Sandwich. Nearby is a village called Ham. Which means that you get this street sign calling out “Ham Sandwich”.
There’s also a bus that starts at Samdwich, goes through Ham, before finishing at the coastal town of Deal.
English, eh?
I was on Mike's Gaming Gala, tonight, talking about the #Intellivision version of #FujiNet https://www.youtube.com/watch?v=BmIEeO1y1O0
A new #FreePascal lesson, thanks to #CompilerExplorer.
Incrementing a number by an ordinal can be done in a few ways in Free Pascal, but they result in a different number of lines of x86_64 assembly.
Let's say we want to increment i by 1.
PASCAL:
i := i + 1; // ... and ...
i += 1; // produce identical assembly
ASSEMBLY:
movl -16(%rbp),%eax
leal 1(%eax),%eax
movl %eax,-16(%rbp)
However...
PASCAL:
Inc(i);
ASSEMBLY:
addl $1,-16(%rbp)
Granted, this might be before any optimisation takes place, but it's still very interesting to see!
I hadn't realised before today that Inc() will work with numbers other than one. So...
PASCAL:
Inc(i, 2);
ASSEMBLY:
addl $2,-16(%rbp)
It will also work with variables.
Let's say we want to increment i by AInt, an incoming argument.
PASCAL:
i := i + AInt; // ... and ...
i += AInt; // produce identical assembly
ASSEMBLY:
movl -16(%rbp),%eax
movl -8(%rbp),%edx
leal (%eax,%edx),%eax
movl %eax,-16(%rbp)
However...
PASCAL:
Inc(i, AInt);
ASSEMBLY:
movl -8(%rbp),%eax
addl %eax,-16(%rbp)
RE: https://mastodon.social/@devilgirlbot/117119504785240510
THE WAR OF THE GARGANTUAS!
I really wish I understood how this Monsterdon thing works because I am totally all over The War of The Gargantuas.
Three words: Rooftop Musical Interlude.
When DEVO covers an incredibly obscure song and it's nowhere near as baffling and demented as the original, you know you have a winner.
I cannot be trusted with movies - this is all the evidence you need.
Relatable Patent Art.
US12711711-B1: The illustrations in this patent are great, they read like a silent children's book about a smirking creeper.
https://jwz.org/b/yk-x
Well, my recovery of TACT3 is broken and I'm not sure it's worth attempting to fix. On the upside it does seem to read and echo the main scenario input correctly. The problem is in the radioactive decay chains. Only one of about a dozen seems to be read in correctly. Maybe that can be isolated to the preprocessor and inventory library reader. However I had to hack on most of the math routines because of stupid F66 array dimensioning issues.
Most of the problem is due to not having any usable test cases until now. I need to rethink whether this project is worth continuing or if possibly I need to revert the whole repo to square one and see if I can get the tests running with the minimally recovered code (or the original DOS executable under DOSBOX or Wine).
Demoralizing. Maybe I can salvage the input processor and figure out how to replace the ancient LINPACK/EISPACK routines. That needs to happen anyway. Or I could just punt and try reimplementing in Python. That's a waste of time because I'd be debugging new code with no pedigree and at the end I'd have yet another undeployable unmaintainable urine-soaked bag of parts just like every other Python application I build. None of them survive the passage of time and instability of the Python ecosystem. The big selling point of keeping the code in Fortran is that you get a single easily deployable binary that can be reliably regenerated for the forseeable future - it's basically bulletproof for software lifecycle management. I'm confident the code will compile and run in 5 days, 5 weeks, 5 months, or 5 years. I have zero faith Python code will properly rebuild a month from now.
FWIW, last week's trip to Albuquerque was to look at the dev and release process for a very large Fortran code that's been under continuous development since the mid-1980s. It's in pretty decent shape, far better than most codes I work with. Their release process is mature and sensible (GitLab + CI pipelines for Mac, Windows, and Linux). I just wish the other codes owned by the regulator were anywhere near the quality of this code.
joined a call with a vendor 3 minutes early today and just fed the ai assistant that joined details about my chosen pricing structure and how we’d all agreed on it and to ignore the rest of the discussion, they usually send the notes through pretty quickly, not today for some reason
Tonight!
UK Racket meet-up (London) Tuesday 18 August 2026 7:30pm
at The City Pride 🍕 28 Farringdon Ln, London EC1R 3AU
Join us for discussion and pizza. All welcome.😁
*discussion about how to write text emoji*
Me: "A face is not a face without a nose."
Friend: "Tell that to skeletor."
Me: "He has a nose.... hole"
Just a blue state boys
Never meanin' no harm...
Would you like your morning toast 🍞 Art Deco style?
👉 It's the Universal Model E9410 "Sweetheart" toaster, patented and produced by Landers, Frary & Clark in 1929, an iconic Art Deco electric appliance.
More on vintage toasters here:
https://rarehistoricalphotos.com/vintage-toasters/
How did we get from "It’s now safe to turn off your computer" to "It’s not safe to turn on your computer"?
Overcompensating comic from 2010 was pretty upsetting then and is much more upsetting now https://www.wigucomics.com/oc/index.php?comic=1281
RE: https://infosec.exchange/@josephcox/117111490746162211
Name and shame please.
"Trump approval falls to 33%, lowest of his presidency, Reuters/Ipsos poll finds"
What this should tell you is, plainly, that 1 in 3 of your neighbors is a Nazi who means you harm.
An expert witness in a lawsuit a Houston explosion that killed three people used ChatGPT to write significant portions of his “expert report.” Said the report should “show how 3M is 0% at fault for the explosion”
Judge not taking it
It's time for mamdatory training on updated procedures and today's subject is Records Management and Document Control, the fundamental bedrock upon which all nuclear work takes place. "Receipts or It Didn't Happen"
Under general notes, we have a section that requires all records and documents to be submitted in English. If an original document is in another language, it will be completely translated to English before it's admitted to the records management system.
"English translations shall be performed or verified by a qualified human translator."
Then it gets better:
"Machine-generated translations, including translations produced solely through artificial intelligence (AI) or automated translation software, shall not be used as the official controlled version of a document. This ensures the technical accuracy of controlled documents and protects Export Controlled Information (ECI), Confidential Information, and other controlled information from unauthorized disclosure through external translation services."
In short, AI is an unacceptable safety and security risk in our field.
RE: https://mastodon.social/@billbot/116928641845733924
TIL "letter of marque" 👀