Pin LF line endings with a root .gitattributes

.gitattributes

@@ -0,0 +1,33 @@
+# Every text file is checked out with LF, whatever core.autocrlf says. The release publishes
+# byte counts of these files (dojogenesis.com/proof, checked by the website's gen_proof.py
+# --check), bash hooks need LF to run, and a CRLF checkout on Windows made every count read high.
+* text=auto eol=lf
+
+# Named so they are always text, even where detection would guess otherwise.
+*.md    text eol=lf
+*.py    text eol=lf
+*.sh    text eol=lf
+*.json  text eol=lf
+*.jsonl text eol=lf
+*.js    text eol=lf
+*.ts    text eol=lf
+*.tsx   text eol=lf
+*.txt   text eol=lf
+*.yaml  text eol=lf
+*.yml   text eol=lf
+
+# Binary: never converted. None are tracked today; listed so a future one is not mangled.
+*.png   binary
+*.jpg   binary
+*.jpeg  binary
+*.gif   binary
+*.ico   binary
+*.webp  binary
+*.pdf   binary
+*.woff  binary
+*.woff2 binary
+*.ttf   binary
+*.otf   binary
+*.zip   binary
+*.gz    binary
+*.tgz   binary

CHANGELOG.md

@@ -67,6 +67,10 @@ method, the rubric and the tables. It started from five misfires reported from o
   deny. The shell parser records each substitution's parent segment so the guard can tell.
 - **`dojo-gates` 0.2.0, staging** — `git add -N` (intent to add) and `git add -n` (dry run) stage no content and no
   longer deny.
+- **LF on every checkout** — a root `.gitattributes` (`* text=auto eol=lf`, with the common text types named and
+  binaries marked) pins line endings, so a Windows clone with `core.autocrlf=true` gets the same bytes as git
+  stores. Before, such a clone had CRLF files, every byte count on /proof read high, and the website's deploy guard
+  refused the deploy unless `DOJO_PLUGINS_DIR` pointed at an LF export.
 
 ## Unreleased — 2.3.0 — the desktop pass