Drop pstdbool.h and pstdint.h

Shimming compilation environments to pre-C99 is no longer worth the
space it takes up.

Author: hostilefork

CREDITS.md

@@ -121,11 +121,6 @@ not be complete.  Please amend with any corrections.
 - Copyright The Mbed TLS Contributors
 - `%extensions/crypt/mbedtls/*`
 
-**pstdint.h**
-- Copyright (c) 2005-2016 Paul Hsieh
-- BSD License
-- `%src/include/pstdint.h`
-
 **qsort**
 - Copyright (c) 1992, 1993 The Regents of the University of California.
 - `%src/core/f-qsort.c`

src/core/a-constants.c

@@ -44,8 +44,6 @@
 #include <stdlib.h> // size_t and other types used in rebol.h
 #include <assert.h>
 
-#include "pstdint.h" // polyfill <stdint.h> for pre-C99/C++11 compilers
-#include "pstdbool.h" // polyfill <stdbool.h> for pre-C99/C++11 compilers
 #include "c-enhanced.h"
 
 #include "tmp-constants.h" // need the extern definitions

src/include/c-enhanced.h

@@ -38,38 +38,25 @@
 #define C_ENHANCED_H
 
 
-//=//// EXPECTS <stdint.h> OR "pstdint.h" SHIM INCLUDED ///////////////////=//
-//
-// Rebol's initial design targeted C89 and old-ish compilers on a variety of
-// systems.  A comment in that code said:
-//
-//     "One of the biggest flaws in the C language was not
-//      to indicate bitranges of integers. So, we do that here.
-//      You cannot 'abstractly remove' the range of a number.
-//      It is a critical part of its definition."
-//
-// Once C99 arrived, the file <stdint.h> offered several basic types, and
-// basically covered the needs:
-//
-// http://en.cppreference.com/w/c/types/integer
-//
-// The code was changed to use either the C99 types -or- a portable shim that
-// could mimic the types (with the same names) on older compilers.  It should
-// be included before %c-enhanced.h is included.
-//
-//=////////////////////////////////////////////////////////////////////////=//
-//
-// Note: INT32_MAX and INT32_C can be missing in C++ builds on some older
-// compilers without __STDC_LIMIT_MACROS and __STDC_CONSTANT_MACROS:
-//
-// https://sourceware.org/bugzilla/show_bug.cgi?id=15366
+#if !defined(DEBUG_CHECK_OPTIONALS)
+    #define DEBUG_CHECK_OPTIONALS 0
+#endif
+
+#if !defined(DEBUG_CHECK_NEVERNULL)
+    #define DEBUG_CHECK_NEVERNULL 0
+#endif
+
+
+// Ren-C assumes the availability of <stdint.h> and <stdbool.h>
 //
-// You can run into this since pstdint.h falls back on stdint.h if it
-// thinks it can.  Put those on the command line if needed.
+// If for some reason the only barrier to compiling the codebase is lack of
+// these definitions, there are shims on the Internet that implement them
+// (look for pstdint.h and pstdbool.h)  But lack of variadic macros is likely
+// to be a bigger showstopper on older platforms.
 //
-// !!! One aspect of pstdint.h is that it considers 64-bit "optional".
-// Some esoteric platforms may have a more hidden form of 64-bit support,
-// e.g. this case from R3-Alpha for "Windows VC6 nonstandard typing":
+// * One aspect of pstdint.h is that it considers 64-bit "optional".
+//   Some esoteric platforms may have a more hidden form of 64-bit support,
+//   e.g. this case from R3-Alpha for "Windows VC6 nonstandard typing":
 //
 //     #if (defined(_MSC_VER) && (_MSC_VER <= 1200))
 //         typedef _int64 i64;
@@ -78,26 +65,18 @@
 //         #define U64_C(c) c ## U64
 //     #endif
 //
-// If %pstdint.h isn't trying hard enough for an unsupported platform of
-// interest to get 64-bit integers, then patches should be made there.
+//   If %pstdint.h isn't trying hard enough for an unsupported platform of
+//   interest to get 64-bit integers, then patches should be made there.
 //
-
-#if !defined(DEBUG_CHECK_OPTIONALS)
-    #define DEBUG_CHECK_OPTIONALS 0
-#endif
-
-#if !defined(DEBUG_CHECK_NEVERNULL)
-    #define DEBUG_CHECK_NEVERNULL 0
-#endif
-
-
-//=//// EXPECTS <stdbool.h> OR "pstdbool.h" SHIM INCLUDED /////////////////=//
+// * INT32_MAX and INT32_C can be missing in C++ builds on some older
+//   compilers without __STDC_LIMIT_MACROS and __STDC_CONSTANT_MACROS:
 //
-// Historically Rebol used TRUE and FALSE uppercase macros, but so long as
-// C99 has added bool to the language, there's not much point in being
-// compatible with codebases that have `char* true = "Spandau";` or similar
-// in them.  So Rebol can use `true` and `false.
+// https://sourceware.org/bugzilla/show_bug.cgi?id=15366
 //
+#include <stdint.h>
+#if !defined(__cplusplus)
+    #include <stdbool.h>
+#endif
 
 
 //=//// EXPECTS <assert.h> TO BE INCLUDED, PATCH BUG //////////////////////=//
@@ -182,10 +161,12 @@
 
 //=//// CPLUSPLUS_11 PREPROCESSOR DEFINE //////////////////////////////////=//
 //
-// Because the goal of Ren-C is ultimately to be built with C, the C++ build
-// is just for static analysis and debug checks.  This means there's not much
-// value in trying to tailor reduced versions of the checks to old ANSI C++98
-// compilers, so the "C++ build" is an "at least C++11 build".
+// Because the goal of Ren-C is generally to be buildable as C, the C++ build
+// is mostly for static analysis and debug checks.  (But also exceptions with
+// try/catch, on platforms where setjmp()/longjmp() are not viable.)
+//
+// There's not much value in trying to tailor reduced versions of the checks
+// to ANSI C++98 compilers, so the "C++ build" is an "at least C++11 build".
 //
 // Besides being a little less verbose to use, testing via a #define allows
 // override when using with Microsoft Visual Studio via a command line