Replace TRegExpr with purpose-built backtracking bytecode VM regex engine

.github/workflows/ci.yml

@@ -134,7 +134,6 @@ jobs:
           GEN_DIR="$PREFIX/lib/fpc/${FPC_VERSION}/units/${TARGET}/rtl-generics"
           FCL_DIR="$PREFIX/lib/fpc/${FPC_VERSION}/units/${TARGET}/fcl-process"
           FCL_BASE_SRC="$PREFIX/share/fpcsrc/packages/fcl-base/src"
-          REGEXPR_SRC="$PREFIX/share/fpcsrc/packages/regexpr/src"
           FCL_NET_SRC="$PREFIX/share/fpcsrc/packages/fcl-net/src"
           OPENSSL_SRC="$PREFIX/share/fpcsrc/packages/openssl/src"
 
@@ -144,11 +143,9 @@ jobs:
           echo "rtl-generics units: $(ls "$GEN_DIR"/*.ppu 2>/dev/null | wc -l) .ppu files"
           echo "fcl-process units: $(ls "$FCL_DIR"/*.ppu 2>/dev/null | wc -l) .ppu files"
           echo "fcl-base source path: $FCL_BASE_SRC"
-          echo "regexpr source path: $REGEXPR_SRC"
           echo "fcl-net source path: $FCL_NET_SRC"
           echo "openssl source path: $OPENSSL_SRC"
           test -d "$FCL_BASE_SRC"
-          test -d "$REGEXPR_SRC"
           test -d "$FCL_NET_SRC"
           test -d "$OPENSSL_SRC"
 
@@ -168,7 +165,7 @@ jobs:
             "$CROSS_FPC" -T"${OS}" -O4 -dPRODUCTION -Xs -CX -XX -B \
               -Fu./source/units -Fu./source/generated -Fu./source/shared -Fu./source/app -Fu./souffle \
               -Fi./source/units -Fi./source/shared -Fi./souffle \
-              -Fu"$RTL_DIR" -Fu"$OBJPAS_DIR" -Fu"$GEN_DIR" -Fu"$FCL_DIR" -Fu"$FCL_BASE_SRC" -Fu"$REGEXPR_SRC" -Fu"$FCL_NET_SRC" -Fu"$OPENSSL_SRC" \
+              -Fu"$RTL_DIR" -Fu"$OBJPAS_DIR" -Fu"$GEN_DIR" -Fu"$FCL_DIR" -Fu"$FCL_BASE_SRC" -Fu"$FCL_NET_SRC" -Fu"$OPENSSL_SRC" \
               -FU"build/compiled" -FE"build" \
               $EXTRA_FLAGS \
               -dFPC_SOFT_FPUX80 \
@@ -184,7 +181,7 @@ jobs:
             "$CROSS_FPC" -T"${OS}" -O4 -dPRODUCTION -Xs -CX -XX -B \
               -Fu./source/units -Fu./source/generated -Fu./source/shared -Fu./source/app -Fu./souffle \
               -Fi./source/units -Fi./source/shared -Fi./souffle \
-              -Fu"$RTL_DIR" -Fu"$OBJPAS_DIR" -Fu"$GEN_DIR" -Fu"$FCL_DIR" -Fu"$FCL_BASE_SRC" -Fu"$REGEXPR_SRC" -Fu"$FCL_NET_SRC" -Fu"$OPENSSL_SRC" \
+              -Fu"$RTL_DIR" -Fu"$OBJPAS_DIR" -Fu"$GEN_DIR" -Fu"$FCL_DIR" -Fu"$FCL_BASE_SRC" -Fu"$FCL_NET_SRC" -Fu"$OPENSSL_SRC" \
               -FU"build/compiled" -FE"build" \
               $EXTRA_FLAGS \
               -dFPC_SOFT_FPUX80 \
@@ -197,7 +194,7 @@ jobs:
           "$CROSS_FPC" -T"${OS}" -O4 -dPRODUCTION -Xs -CX -XX -B \
             -Fu./source/units -Fu./source/generated -Fu./source/shared -Fu./source/app -Fu./souffle \
             -Fi./source/units -Fi./source/shared -Fi./souffle \
-            -Fu"$RTL_DIR" -Fu"$OBJPAS_DIR" -Fu"$GEN_DIR" -Fu"$FCL_DIR" -Fu"$FCL_BASE_SRC" -Fu"$REGEXPR_SRC" -Fu"$FCL_NET_SRC" -Fu"$OPENSSL_SRC" \
+            -Fu"$RTL_DIR" -Fu"$OBJPAS_DIR" -Fu"$GEN_DIR" -Fu"$FCL_DIR" -Fu"$FCL_BASE_SRC" -Fu"$FCL_NET_SRC" -Fu"$OPENSSL_SRC" \
             -FU"build/compiled" -FE"build" \
             $EXTRA_FLAGS \
             -dFPC_SOFT_FPUX80 \
@@ -209,7 +206,7 @@ jobs:
           "$CROSS_FPC" -T"${OS}" -O4 -dPRODUCTION -Xs -CX -XX -B \
             -Fu./source/units -Fu./source/generated -Fu./source/shared -Fu./source/app -Fu./souffle \
             -Fi./source/units -Fi./source/shared -Fi./souffle \
-            -Fu"$RTL_DIR" -Fu"$OBJPAS_DIR" -Fu"$GEN_DIR" -Fu"$FCL_DIR" -Fu"$FCL_BASE_SRC" -Fu"$REGEXPR_SRC" -Fu"$FCL_NET_SRC" -Fu"$OPENSSL_SRC" \
+            -Fu"$RTL_DIR" -Fu"$OBJPAS_DIR" -Fu"$GEN_DIR" -Fu"$FCL_DIR" -Fu"$FCL_BASE_SRC" -Fu"$FCL_NET_SRC" -Fu"$OPENSSL_SRC" \
             -FU"build/compiled" -FE"build" \
             $EXTRA_FLAGS \
             -dFPC_SOFT_FPUX80 \

.github/workflows/toolchain.yml

@@ -107,9 +107,8 @@ jobs:
 
           # Keep official package sources available for cross builds. The
           # cached cross toolchain only prebuilds a minimal package subset, so
-          # source-based lookup is needed for units like Base64 and RegExpr.
+          # source-based lookup is needed for units like Base64.
           cp -R "$GITHUB_WORKSPACE/fpc-source/packages/fcl-base" "$PREFIX/share/fpcsrc/packages/"
-          cp -R "$GITHUB_WORKSPACE/fpc-source/packages/regexpr" "$PREFIX/share/fpcsrc/packages/"
           cp -R "$GITHUB_WORKSPACE/fpc-source/packages/fcl-net" "$PREFIX/share/fpcsrc/packages/"
           cp -R "$GITHUB_WORKSPACE/fpc-source/packages/openssl" "$PREFIX/share/fpcsrc/packages/"
 

docs/build-system.md

@@ -433,7 +433,7 @@ It:
 6. If the `FPC_TARGET_CPU` environment variable is set, prepends `-P<arch>` to the compiler arguments (used by CI to target x86_64 on Windows where the FPC package defaults to i386).
 7. For the `tests` target, auto-discovers all `*.Test.pas` files in `source/units/` and `source/shared/`.
 
-The GitHub Actions cross-compilation workflow uses a reduced cached FPC toolchain rather than a full target-side FCL install. It prebuilds the RTL, `rtl-objpas`, `rtl-generics`, and `fcl-process`, and also caches the official `fcl-base` and `regexpr` sources so cross builds can resolve units such as `Base64` and `RegExpr` on demand from the shipped FPC packages.
+The GitHub Actions cross-compilation workflow uses a reduced cached FPC toolchain rather than a full target-side FCL install. It prebuilds the RTL, `rtl-objpas`, `rtl-generics`, and `fcl-process`, and also caches the official `fcl-base` sources so cross builds can resolve units such as `Base64` on demand from the shipped FPC packages.
 
 ## Project Structure for Compilation
 

docs/built-ins.md

@@ -236,7 +236,7 @@ RegExp is available as both `RegExp()` and `new RegExp()`. Regex literals (`/pat
 - When the replacer is a function and named groups are present, the `groups` object is passed as the last argument after `input`.
 - `String.prototype.match`, `matchAll`, `replace`, `replaceAll`, `search`, and `split` dispatch through the corresponding well-known symbol hooks, so custom protocol objects work as expected.
 - `matchAll()` returns a lazy iterator that advances matches on demand per the specification.
-- The `u` flag enables Unicode-aware pattern matching. Unicode property escapes (`\p{Letter}`, `\P{ASCII}`, etc.) are expanded to equivalent character classes. Unicode code point escapes (`\u{41}`, `\u{1F600}`) are converted to UTF-8 byte sequences. Supported properties: `L`/`Letter`, `Lu`/`Uppercase_Letter`, `Ll`/`Lowercase_Letter`, `N`/`Number`, `Nd`/`Decimal_Number`, `P`/`Punctuation`, `S`/`Symbol`, `Z`/`Separator`, `Cc`/`Control`, `ASCII`, `ASCII_Hex_Digit`, `White_Space`. Unsupported properties throw `SyntaxError`. The `u` flag also disables TRegExpr's Russian charset extensions and enables correct `AdvanceStringIndex` for multi-byte UTF-8 sequences.
+- The `u` flag enables Unicode-aware pattern matching. Unicode property escapes (`\p{Letter}`, `\P{ASCII}`, etc.) are matched against Unicode code point range tables. Unicode code point escapes (`\u{41}`, `\u{1F600}`) are converted to UTF-8 byte sequences. Supported properties: `L`/`Letter`, `Lu`/`Uppercase_Letter`, `Ll`/`Lowercase_Letter`, `N`/`Number`, `Nd`/`Decimal_Number`, `P`/`Punctuation`, `S`/`Symbol`, `Z`/`Separator`, `Cc`/`Control`, `ASCII`, `ASCII_Hex_Digit`, `White_Space`. Unsupported properties throw `SyntaxError`. The `u` flag enables correct `AdvanceStringIndex` for multi-byte UTF-8 sequences.
 - The `v` flag (Unicode sets) is accepted and exposed through `.flags` and `.unicodeSets`. The `u` and `v` flags are mutually exclusive. Full Unicode set notation and properties of strings in character classes are not yet implemented beyond basic `u` flag behavior.
 - The `d` flag (indices) is accepted and exposed through `.flags` and `.hasIndices`. Match indices are not yet populated.
 

docs/decision-log.md

@@ -17,6 +17,8 @@ Chronological record of key architectural and implementation decisions, newest f
 
 ---
 
+**2026-05-08** · `engine` — Replace TRegExpr with a purpose-built backtracking bytecode VM regex engine. TRegExpr used native call recursion for backtracking, causing SIGSEGV on inputs ~42K+ chars when combined with the evaluator's stack depth (#515). Three preprocessing passes papered over feature gaps: `(?s)` modifier scope leak, no named groups (two-pass rewrite), and inadequate Unicode (`\p{...}` expanded to ASCII approximations). New architecture: `Goccia.RegExp.Compiler.pas` (recursive-descent parser + bytecode emitter) and `Goccia.RegExp.VM.pas` (iterative dispatch loop with heap-allocated backtrack stack and always-on failure memoization). The compiler parses ES2026 regex grammar directly, handling named groups, Unicode property escapes, inline modifier groups, and backreferences natively — no preprocessing passes. The memoization cache records `(PC, InputPos)` failure states to prune exponential backtracking (e.g., `(a+)+b`). Configurable step limit (default 10M) throws `Error` instead of crashing. Removes the FPC `regexpr` package from the cross-compilation toolchain. Reuses `TextSemantics.pas` UTF-8 functions (`TryReadUTF8CodePoint`, `AdvanceUTF8StringIndex`, `CodePointToUTF8`, etc.) rather than reimplementing. Public API (`ExecuteRegExp` signature, `TGocciaRegExpMatchResult` record) unchanged; `Goccia.RegExp.Runtime.pas` and `Goccia.Builtins.GlobalRegExp.pas` unmodified.
+
 **2026-05-05** · `parser` — Opt-in traditional `for(init; test; update)` loops (`--compat-traditional-for-loop`). Added behind a new compatibility flag for ECMAScript compatibility when porting legacy code, mirroring the existing `--compat-var` and `--compat-function` posture. The flag is off by default and ORed in by `--compat-all` so test262 (which always passes `--compat-all`) executes traditional `for(;;)` bodies that previously parser-warn-and-skipped — surfacing real engine gaps in unrelated areas (Atomics #541, Intl #542, BigInt postfix increment #540, etc.). `let`/`const` declarations in for-init create a per-iteration lexical environment per ES2026 §14.7.4.4, so closures captured during iteration N pin to that iteration's binding (the textbook `fns.push(() => i)` case yields `[0, 1, 2]`, not `[3, 3, 3]`). `var` declarations require both `--compat-var` and the new flag and share a single hoisted binding visible after the loop. The bytecode compiler reuses the counted-loop pattern from `CompileCountedForOf` for `for(let i = N; i <op> M; i++ | i--)` shapes (rejecting var/const, bodies that mutate the loop var, type annotations, and non-integer-literal cond RHS). `while` and `do...while` remain excluded — they have the same stub status but were intentionally split into a separate iteration. [language-tables.md](language-tables.md).
 
 **2026-05-04** · `testing` · [#513](https://github.com/frostney/GocciaScript/pull/513) — test262 conformance harness reframed around the standard tc39 convention. Previously the wrapper ran inside `GocciaTestRunner` and had to selectively hide / capture / restore the test-library globals (`expect`, `describe`, `test`, `runTests`, etc.) it registered, with failure capture leaning on an `undefined` sentinel that collided with thrown `undefined` and chunked-runner crashes that masked thousands of conformance failures as wrapper failures (#491 history). Replaced with: per-test `GocciaScriptLoaderBare` subprocess, stock tc39/test262 harness files read directly from the pinned checkout's `harness/` directory (with a small set of bundled adaptations under `scripts/test262_harness/` for stock files that depend on language features Goccia excludes by design or that work around specific engine bugs — see [test262.md § Bundled harness adaptations](test262.md#bundled-harness-adaptations)), exit-code + stdout-marker wire protocol identical to `test262-harness`/`eshost`/test262.fyi, and a thin TypeScript orchestrator (`scripts/run_test262_suite.ts`). No eligibility filter — every discovered test runs; per-test subprocess + `--timeout` + `--max-memory` bound the blast radius. Wrapper-template drift is now structurally impossible because the "template" is `harness + body` string concatenation. Wrapper-infra failures are classified separately and gated to zero in CI. Surfaced eleven engine bugs (all milestoned 0.8.0, all labeled `engine` per the architecture split where `engine` covers `TGocciaEngine` — language semantics + ECMAScript built-ins — and `runtime` is reserved for `TGocciaRuntime` host extensions like console/fetch/JSON5): [#514](https://github.com/frostney/GocciaScript/issues/514) (Iterator.concat SIGSEGV), [#515](https://github.com/frostney/GocciaScript/issues/515) (RegExp.test SIGSEGV), [#516](https://github.com/frostney/GocciaScript/issues/516) (`Reflect.construct` rejects function decls/exprs), [#517](https://github.com/frostney/GocciaScript/issues/517) (script-mode unattached call `this`), [#518](https://github.com/frostney/GocciaScript/issues/518) (bytecode VM Range-check on top-level `Promise.then` drain), [#519](https://github.com/frostney/GocciaScript/issues/519) (`Error.prototype.constructor` missing), [#520](https://github.com/frostney/GocciaScript/issues/520) (module arrow `this` lexical inheritance), [#521](https://github.com/frostney/GocciaScript/issues/521) (`var`/`function` shadowing built-in globals), [#522](https://github.com/frostney/GocciaScript/issues/522) (`String(obj)` doesn't invoke `toString`), [#523](https://github.com/frostney/GocciaScript/issues/523) (`yield*` accesses `.next` on null), [#524](https://github.com/frostney/GocciaScript/issues/524) (for-of re-fetches `iterator.next` each iteration). Each bundled-harness adaptation under `scripts/test262_harness/` references its tracking issue and is to be removed when the underlying engine bug is fixed. [test262.md](test262.md).

scripts/run_test262_suite.ts

@@ -74,8 +74,6 @@ const DEFAULT_JOBS = 4;
 // bug.  Per docs/test262.md "Updating the contract", this list is the
 // only allowed form of test-skipping; no generic eligibility filter.
 const KNOWN_ENGINE_CRASHES = new Set<string>([
-  // SIGSEGV: RegExp.prototype.test trailing-input edge case. https://github.com/frostney/GocciaScript/issues/515
-  "staging/sm/RegExp/test-trailing.js",
 ]);
 
 // ---------------------------------------------------------------------------