Bring regexx.c/regexx.h's own file-header comments up to date

Both still described the v1 implementation as a single backtracking
engine with no mention of the Pike VM at all, stale since the engine
was added several commits ago; regexx.c's went further and never
mentioned it even in passing. Updated both to describe the actual
dual-engine dispatch, the Pike VM's allocation-free thread model and
literal prefilter (concept.md 7.10), and corrected the reentrancy note
from a blanket "do not call from more than one thread" to the actual,
narrower guarantee this session's PikeEngine design specifically
preserved: re_compile and the module level cache still need external
synchronization across threads, but Pattern_-prefixed calls on an
already-compiled Pattern do not, because nothing mutates a compiled
Pattern and PikeEngine is always scoped to one top-level call, never
cached on the Pattern itself.

Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01EjuMk8kY9SDus1wWe2K9xY

regexx.c

@@ -2,29 +2,45 @@
  *
  * See concept.md for the full design rationale. This file is the v1
  * implementation: it covers the public API and pattern syntax listed in
- * README.md "Implementation status" in full, executed by a single
- * recursive backtracking engine (concept.md Section 7.3) over a fully
- * materialized copy of the input. It does not yet implement the
- * streaming, bounded-memory regular engine of concept.md Section 7.2;
- * that remains future work, tracked in README.md.
+ * README.md "Implementation status" in full, executed by one of two
+ * engines over a fully materialized copy of the input, chosen
+ * automatically per compiled pattern (concept.md Section 7.2/7.7):
  *
- * That backtracking engine is memoized (run_memo) and, for quantifiers
- * over a single character/class/./ (OP_REPEAT1), backed by precomputed
- * run-length and skip-ahead tables (compute_maxrun,
- * compute_next_prevmatch), so that Pattern_search/finditer/split/sub
- * run in linear rather than quadratic time for the common,
- * backreference-free case; see README.md "Implementation status" for
- * the measured numbers, the precise remaining limitations, and
- * citations to the published techniques this matches.
+ * - A Pike VM (Section 6b, "6b. Pike VM" below), a Thompson-NFA
+ *   simulation with no backtracking at all, for any pattern with no
+ *   backreference, lookahead, lookbehind, or atomic group. No
+ *   per-thread heap allocation (a flat table indexed directly by
+ *   instruction, concept.md 7.10) and a Boyer-Moore-Horspool literal
+ *   prefilter for a pattern beginning with a mandatory literal string.
+ * - A recursive backtracking engine (Section 6, concept.md 7.3) for
+ *   everything else, memoized (run_memo) and, for quantifiers over a
+ *   single character/class/./ (OP_REPEAT1), backed by precomputed
+ *   run-length and skip-ahead tables (compute_maxrun,
+ *   compute_next_prevmatch), so that Pattern_search/finditer/split/sub
+ *   run in linear rather than quadratic time for the common,
+ *   backreference-free case there too.
+ *
+ * Neither engine yet implements the streaming, bounded-memory regular
+ * engine of concept.md Section 7.2 (both still materialize the whole
+ * input); that remains future work, tracked in README.md. See
+ * README.md "Implementation status" and "The Pike VM" for the measured
+ * numbers on both engines, the precise remaining limitations, and
+ * citations to the published techniques and production engines
+ * (RE2, rust-lang/regex) this matches.
  *
  * Naming convention (concept.md Section 9.0): every identifier with a
  * direct Python `re` counterpart uses that counterpart's exact spelling.
  *
- * Not reentrant: re_compile() uses one process-wide parser scratch pool
+ * Reentrancy: re_compile() uses one process-wide parser scratch pool
  * and the module level functions share one process-wide pattern cache
- * (concept.md 13.6); do not call regexx functions from more than one
- * thread without external synchronization.
- */
+ * (concept.md 13.6); do not call re_compile() or any re_-prefixed
+ * module level function from more than one thread without external
+ * synchronization. Pattern_-prefixed functions on an already-compiled
+ * Pattern no thread is concurrently modifying need no such
+ * synchronization (nothing here mutates a compiled Pattern, including
+ * matching on the Pike VM: its per-search working state, PikeEngine,
+ * is always local to one top level call, never cached on the Pattern
+ * itself, specifically so this guarantee holds, concept.md 7.10). */
 #define _DEFAULT_SOURCE
 #include "regexx.h"
 #include <stdlib.h>

regexx.h

@@ -9,19 +9,21 @@
  * described below, executed by one of two engines over a fully
  * materialized copy of the input, chosen automatically per compiled
  * pattern (concept.md Section 7.2/7.7): a Pike VM (a Thompson-NFA
- * simulation, genuinely linear time, no backtracking at all) for any
- * pattern with no backreference, lookaround, or atomic group; a
- * recursive backtracking engine (concept.md Section 7.3), augmented
- * with memoization and precomputed skip-ahead tables (regexx.c:
- * run_memo, compute_maxrun, compute_next_prevmatch) that make the
- * search-family functions (Pattern_search/finditer/split/sub) run in
- * linear time for the common, backreference-free case there too,
- * otherwise. Neither engine yet implements the streaming, bounded-
- * memory version of Section 7.2; see README.md "Implementation status"
- * and "The Pike VM" for the complete, itemized, measured account of
- * what v1 does and does not cover, including a real, documented
- * performance trade-off between the two engines that is not yet
- * resolved (concept.md 7.8).
+ * simulation, genuinely linear time, no backtracking at all, no
+ * per-thread heap allocation, and a Boyer-Moore-Horspool literal
+ * prefilter, concept.md 7.10) for any pattern with no backreference,
+ * lookaround, or atomic group; a recursive backtracking engine
+ * (concept.md Section 7.3), augmented with memoization and precomputed
+ * skip-ahead tables (regexx.c: run_memo, compute_maxrun,
+ * compute_next_prevmatch) that make the search-family functions
+ * (Pattern_search/finditer/split/sub) run in linear time for the
+ * common, backreference-free case there too, otherwise. Neither engine
+ * yet implements the streaming, bounded-memory version of Section 7.2;
+ * see README.md "Implementation status" and "The Pike VM" for the
+ * complete, itemized, measured account of what v1 does and does not
+ * cover, including the remaining, narrower performance trade-off
+ * between the two engines on patterns the Pike VM's prefilter cannot
+ * help (concept.md 7.9/7.10).
  */
 #ifndef REGEXX_H
 #define REGEXX_H