From 4efa1f41503e0bd2e8c4ce66903903a7c1d117fd Mon Sep 17 00:00:00 2001 From: Rafael Vuijk Date: Tue, 27 Oct 2026 13:58:23 +0000 Subject: [PATCH] A fix for an approximation quirk before v3 links the decision that removes it On #2139 the maintainer asked that fixes for approximation quirks made before v3 be simplified away for v3. v3 removes downcasting (#1019, comment 5754734225), and with it the behaviour those fixes work around. For that to happen, the fixes have to be findable, so AGENTS.md asks each one to link the decision in its comment and in its test. A search for the comment id then lists them. Co-Authored-By: Claude Opus 5.5 (1M context) Claude-Session: https://claude.ai/code/session_012sonx8iAspMiwRwokT1Ura --- AGENTS.md | 8 ++++++++ 1 file changed, 8 insertions(+) diff --git a/AGENTS.md b/AGENTS.md index 4ae73fdf4..1af2d72ae 100644 --- a/AGENTS.md +++ b/AGENTS.md @@ -55,6 +55,14 @@ wherever the input was exact, and interval arithmetic for what is genuinely appr shape of the answer, and they are queued for v3 in [#1019](https://github.com/asc-community/AngouriMath/issues/1019). +**A fix for an approximation quirk before v3 is temporary, and it says so.** v3 removes +downcasting ([#1019](https://github.com/asc-community/AngouriMath/issues/1019#issuecomment-5754734225)). +With it go the decimal answers, the rounding onto rationals and the residuals taken as zero that +such a fix works around. These fixes are to be simplified away for v3 +([#2139](https://github.com/asc-community/AngouriMath/issues/2139#issuecomment-6443699564)). So +each one links that decision in its comment and in the test that pins it, and a search for +`5754734225` lists what v3 has to take back out. + ## Not answering is a legitimate answer. Answering wrongly is not. The most important distinction in this codebase: