Dev log: implementing new syntax in GHC

In September 2024, I started working on a patch to GHC. The story starts slightly before that. In July 2024 I wondered Could Undecidable Instances be per-instance?. It seemed like a fairly small improvement that I could possibly even implement myself. The answer was, roughly, "someone's suggested it, and people are broadly in favor, but with different syntax that doesn't yet exist". So it seems the story is, people want to phase out use of these annotations (which are weird because comments shouldn't change whether a program is accepted or not) in favor of extending use of modifier syntax (currently used for linear types, e.g. a %One -> b). But nailing down the exact details of the new syntax is kind of awkward. (Is a semicolon required in such-and-such position? How are competing modifiers handled? Is a modifier a type or a type constructor or what? And something I didn't see discussed: if I do type MySynonym = ExistingModifier, can I use %MySynonym in the same module? If so, does that need extra work or does it come for free?) So instead of what seems like a fairly-small change, it's a lot more ambitious. Definitely no longer something that makes me think "oh yeah, I could do that myself". Then a few months later I decided to try implementing the new syntax anyway1. I opened a PR2 in September 2025, and it got merged in April 2026, so about a year and a half after I started. For most of that period I wasn't putting many hours into it. This is how it went. I didn't start writing until January 2026, and I got distracted and didn't finish writing until June 2026, so there's gonna be a bunch of detail missing or misremembered. There's no particular theme or narrative here, it just seemed like a useful kind of thing to have written down. ghc The first step was to get GHC's master branch compiling. This was my second contribution to GHC, so I'd done it before, and it was easy enough. (Previously RyanGlScott had walked me through the process at Zurihac 2023.) The actual e