Ask the code gate which role came back unread, not what ratio it read at

tests/render_cases_layout.py

@@ -623,16 +623,18 @@ def ceiling(limit, approvals):
 UNANSWERED_CODE_PAGE = COLORED_CODE_PAGE.replace(
     "</head>", "<style>[data-lf-syn] { color: inherit; }</style>\n</head>"
 )
-# What --syn-comment carried until this gate was written: 3.3:1 on --pre-bg, and the
-# reading a user reported as the highlighting being gone.
+# What --syn-comment carried until this gate was written: under 4.5:1 on --pre-bg, and
+# the reading a user reported as the highlighting being gone. The colour is stated and
+# the ratio is not, because the ratio is a reading of whatever --pre-bg currently is.
 FAINT_CODE_PAGE = COLORED_CODE_PAGE.replace(
     "</head>", '<style>[data-lf-syn="cm"] { color: #8b8577; }</style>\n</head>'
 )
 
 # A role that reads on the block and not on the tint one of its lines wears. The clean
-# line comes first on purpose: a gate that stopped at a role's first span would take the
-# 7.9:1 reading and never reach the 1.6:1 one two lines down, and a walkthrough's hi band
-# is the surface where a code line is most often set on something other than --pre-bg.
+# line comes first on purpose: a gate that stopped at a role's first span would take that
+# line's reading, which clears the threshold, and never reach the one two lines down, and
+# a walkthrough's hi band is the surface where a code line is most often set on something
+# other than --pre-bg.
 TINTED_LINE_PAGE = LONG_PAGE.replace(
     "</head>", "<style>:root { --hi-tint: #6f6a60; }</style>\n</head>"
 ).replace(

tests/test_render_gate.py

@@ -1874,21 +1874,23 @@ def test_the_render_gate_reports_code_the_reader_cannot_tell_from_its_block(
         if f.startswith("[light] code marked cm is the ink of the code around it")
     ], unanswered
 
+    # The ratio the gate prints is a reading of the theme's own surfaces, so pinning
+    # its digits here makes every palette change a failure of this test rather than of
+    # the page. What the assertions ask instead is which role came back unread, which
+    # is the whole of what each case is arranged to distinguish: the gate names a role
+    # in this sentence only where it read under 4.5:1, so the finding is the claim.
     faint = render_gate_model.render_version(browser, serve(FAINT_CODE_PAGE))
-    assert [
-        f for f in faint if f.startswith("[light] code marked cm reads at 3.3:1")
-    ], faint
+    assert [f for f in faint if f.startswith("[light] code marked cm reads at ")], faint
     assert not [f for f in faint if "code marked st" in f], (
         "only the role the style touched is unread, so the rest name the reading "
         "rather than the rule"
     )
 
     tinted = render_gate_model.render_version(browser, serve(TINTED_LINE_PAGE))
-    assert [
-        f for f in tinted if f.startswith("[light] code marked st reads at 1.6:1")
-    ], (
+    assert [f for f in tinted if f.startswith("[light] code marked st reads at ")], (
         "the reading is of the surface each span is actually set on, not of one "
-        f"block colour taken once per role — {tinted}"
+        "block colour taken once per role — that role clears the threshold on the "
+        f"block, so a gate stopping at its clean line says nothing at all: {tinted}"
     )
 
     page, errors = open_page(browser, serve(SHADOW_CODE_PAGE))