Fix parsing for Sentinel-Terminated Exprs (#32)

* fix parsing for some files

* tweak parsing script

* more explicit message

* fix parsing for for sentinel-terminated exprs

* simplify implementation

grammar.js

@@ -530,7 +530,7 @@ module.exports = grammar({
         // *** Helper grammar ***
         BreakLabel: ($) => seq(COLON, $.IDENTIFIER),
 
-        BlockLabel: ($) => prec.left(seq($.IDENTIFIER, COLON)),
+        BlockLabel: ($) => prec.left(seq(...blockLabel($))),
 
         FieldInit: ($) =>
             seq(DOT, field("field_member", $.IDENTIFIER), EQUAL, $._Expr),
@@ -710,12 +710,29 @@ module.exports = grammar({
                 $.ArrayTypeStart
             ),
 
+        /*
+          Given a sentinel-terminated expression, e.g. `foo[bar..bar: 0]`, note
+          how "bar: " resembles the opening of a labeled block; due to that
+          label-like syntax, Tree-sitter would try to parse what comes after the
+          COLON as a block, which would obviously fail in this case. To work
+          around that problem, we'll create a SentinelTerminatedExpr rule which
+          starts like the the BlockLabel rule (hence why they share a common
+          definition), but ends with an arbitrary Expr instead of a block.
+          BlockLabel should have preferential precedence based on its usage sites
+          throughout the rest of the grammar, thus this rule effectively serves
+          as a fallback for the former.
+        */
+        _SentinelTerminatedExpr: ($) => choice(
+          seq(...blockLabel($), $._Expr),
+          seq($._Expr, optional(seq(COLON, $._Expr)))
+        ),
+
         SuffixOp: ($) =>
             choice(
                 seq(
                     LBRACKET,
                     $._Expr,
-                    optional(seq(DOT2, optional($._Expr), optional(seq(COLON, $._Expr)))),
+                    optional(seq(DOT2, optional($._SentinelTerminatedExpr))),
                     RBRACKET
                 ),
                 DOTASTERISK,
@@ -869,3 +886,12 @@ function keyword(rule, _) {
 function sepBy(sep, rule) {
     return optional(sepBy1(sep, rule));
 }
+
+/*
+  This rule was extracted as a function for the sake of making
+  _SentinelTerminatedExpr and BlockLabel share the same definition. Please check
+  the comment of _SentinelTerminatedExpr for more context.
+*/
+function blockLabel($) {
+  return [$.IDENTIFIER, COLON]
+}

scripts/known-parsing-failures.txt

@@ -1,4 +1,3 @@
-zig/test/cases/safety/slice with sentinel out of bounds - runtime len.zig
 zig/test/behavior/struct.zig
 zig/test/behavior/slice.zig
 zig/test/behavior/array.zig
@@ -10,7 +9,6 @@ zig/lib/std/os.zig
 zig/lib/std/fs.zig
 zig/lib/std/mem.zig
 zig/lib/std/zig/c_translation.zig
-zig/lib/std/zig/system/windows.zig
 zig/lib/std/os/test.zig
 zig/lib/std/os/linux/bpf/btf.zig
 zig/lib/std/os/linux/bpf.zig
@@ -24,19 +22,11 @@ zig/lib/std/os/uefi/protocols/edid_override_protocol.zig
 zig/lib/std/os/uefi/tables/boot_services.zig
 zig/lib/std/unicode.zig
 zig/lib/std/Thread.zig
-zig/lib/std/mem/Allocator.zig
 zig/lib/std/json/test.zig
-zig/lib/std/child_process.zig
 zig/lib/std/macho.zig
-zig/lib/std/process.zig
-zig/lib/std/dwarf.zig
-zig/lib/std/json.zig
 zig/lib/std/c/openbsd.zig
-zig/src/main.zig
 zig/src/link/Wasm/Object.zig
 zig/src/link/Wasm/Archive.zig
 zig/src/link/MachO/Archive.zig
-zig/src/Air.zig
 zig/src/codegen/llvm/bindings.zig
 zig/src/Autodoc.zig
-zig/src/Zir.zig