Notices by Andy Wingo (wingo@mastodon.social), page 3

in c i mostly write compiler runtimes, so i -fno-strict-aliasing usually and usually have some memory safety story already. but hoo boy, integers terrify me

- if a function takes two integers as args, it is very easy to accidentally swap them

i should probably elaborate for people that have not had the "pleasure" of c.
- signed integer overflow is UB. you have to prevent it somehow.
- you always run into funny behavior on the boundaries, e.g. loading a 4-byte value at UINT32_MAX-2 needs to trap, but it's easy to accidentally write a bounds check that wraps
- there is a thicket of annoyance around size conversions, easy to accidentally get sign extension bugs
- approximately every integer error is exploitable

@dotstdy @whitequark the answer is https://ometer.com/preferences.html. was a big vibe 20yo

@dotstdy @whitequark one guy, yes, but a lot of people took it to heart, it wasn’t imposed. different projects will have different ideas of what are essential preferences but everybody still has to draw a line somewhere

@whitequark (llvm.wasm)

@whitequark just testing a wasm-to-c research compiler, https://codeberg.org/andywingo/wastrel

i have a c file generated from wasm that takes 70 minutes to compile at -O1, to generate an 80 MB binary. i wish gcc could parallelise internally (as it can do in an LTO context)

@whitequark catherine it is generated from your wasm file

i tried using C11 _Generic recently. boy was that a mistake https://www.chiark.greenend.org.uk/~sgtatham/quasiblog/c11-generic/

make it stop

great post on what it's like working with a tracing jit in practice https://antocuni.eu/2025/09/24/tracing-jits-in-the-real-world--cpython-core-dev-sprint/#

python types are amazing because they compile down to nothing: you can work on them for couple hours, and the result is no change in behavior, neither at run-time nor at compile-time

@mayintoronto TRAINS

software engineering is good and all, but i just had a python linter tell me i needed more whitespace on a comment that i had to add to disable python type checking for a mock argument that was part of a test suite of a test suite (yes). make it stop

@whitequark in my mind this is leading to you writing a fluid dynamics simulator

"we don't need types, we have types at home" smh, what a notation

rust is just a conspiracy by big serde to sell more serde

i have seen a few people arguing for xslt recently; weird proposition!
- libs have a cost (browser download size, update work, integration, security risk)
- xslt 1.0 is not good. https://okmij.org/ftp/papers/SXSLT-talk.pdf
- nobody uses it. (or rather, 0.001% of page-loads use it.)
- regardless of your pov on xml, xml web documents are entirely a lost cause; the web is html, and html is not xml.
- there are so many better alternatives! even a simple page with a script tag can do a better job!

saw a surprising product in the supermarket today while looking for the butter