Speaking of flash drives... We know that are many cursed cheap-ass USB flash drives that would just brick if you dare to write a non-FAT32 filesystem. Why? There are only guesses: buggy controller logic behaves unpredictably with non-FAT32, journaling increases unclean-unplug FTL corruption risks of already-buggy logic, etc. I wonder if anyone ever seriously engineered or reverse-engineered one of those to see what exactly went wrong in the firmware?

"Strangely the only people who’ve developed software to reset flash controllers are the Russians, maybe they’re used to this. I suppose the Eastern European hackers are famous for really low-level nitty gritty understanding, presumably thanks to a deadly Soviet-era mix of education plus having to make everything work given only a hammer and vodka."Actually more from China for a different reason: it's where they get made... But ultimately most tools are leaks from Taiwanese chipmakers anyway. The leak chain is like: Taiwan -> Shenzhen -> Chinese forums -> Russian forums (sometimes with original contributions from both sides) -> English forums.

Someone should create a "this blog post could be a research paper" award (and perhaps also a "this research paper could be a blog post" award).

Spent half an hour trying to figure out how to escape a semicolon in CMAKE_ARGS. It's a nasty problem, whenever the variable value is used to set another variable, the semicolon character becomes a list separator, so the number of escapes depends on the number of variable expansions... The solution is to use the generator expression $<SEMICOLON>... (only for CMake 3.3+).

@nprofile1q... I didn't know Linux kernel comes with a modern version of the classic IBM PC BIOS memtest.

Lesson for naming features: never use terms such as "new engine" or "fast protocol" as they age poorly. Use codenames.

Q: What plants do cryptographers grow in their backyards?

A: Merkle trees.

#Windows development experience be like... ​:woozy_baa:​

You know the software is from a Big Company when the project doesn't compile with the latest dependency version from the same company.

You must be kidding me. Your CI job failed. Why? No logs, the whole runner just suddenly died. What?!

Pure Mathematicians: No, you can't prove math theorems just by simulating differential equations on a computer. Floating-point operations and discretizations are imprecise.

Applied Mathematicians: Assign an error bar to all math you did, call it a "rigorous“ solver. If the whole bar is in the "correct" region, problem solved.

"autotools: checking whether stack growns downwards... yes"Makes you wonder what happens if it doesn't.