review pass: fix the eaten Tab, drop the duplicated code, cover the gaps

1 files changed, 33 insertions, 0 deletions

diff --git a/test/snapshots/lspcompletemoved.snap b/test/snapshots/lspcompletemoved.snap
new file mode 100644
index 00000000..5f20b806
--- /dev/null
+++ b/test/snapshots/lspcompletemoved.snap
@@ -0,0 +1,33 @@
+# The guard that makes "an empty answer indents" safe to have at all.
+#
+# lspcomplete.snap pins the indent itself: Tab after a dot the backend cannot
+# answer diverts, the answer comes back empty, and four spaces go in LATE so
+# the keystroke is not eaten. That late write is the dangerous half — it lands
+# a few hundred milliseconds after the key, by which time the hands have moved
+# on — so lspResponse only performs it if nothing has: same pane, still insert,
+# one cursor, cursor still on the cell the Tab was pressed at. Nothing covered
+# the condition, only its happy side.
+#
+# Same file and same click as lspcomplete's `commentindent` arm — line 12 is a
+# comment, so no candidate can exist and the answer is certainly empty — but
+# here an `x` is typed straight after the Tab, in the same write, long before a
+# ZLS analysis can finish. The buffer below is the assertion: `// note.x`, with
+# neither `// note.    x` (the indent landing under the typing) nor
+# `// note.x    ` (the indent landing on top of it).
+file a.zig const Shade = enum {\n    verdigris,\n    cinnabar,\n};\n\nfn pick(s: Shade) u8 {\n    return switch (s) {\n        \n    };\n}\n\n// note. see below\n
+start 20 120 a.zig
+wait 8000 verdigris
+stable 700 20000
+press left 16 14
+release left 16 14
+stable 400 5000
+key i
+key tab
+text x
+stable 700 20000
+snap typedthrough
+key esc
+settle 100
+key space f s
+stable 700 15000
+snapfile buffer /tmp/pardes-snap/lspcompletemoved/cwd/a.zig